Dynamics of cell rounding during detachment

Summary Animal cells undergo repeated shape changes, for example by rounding up and respreading as they divide. Cell rounding can be also observed in interphase cells, for example when cancer cells switch from a mesenchymal to an ameboid mode of cell migration. Nevertheless, it remains unclear how interphase cells round up. In this article, we demonstrate that a partial loss of substrate adhesion triggers actomyosin-dependent cortical remodeling and ERM activation, which facilitates further adhesion loss causing cells to round. Although the path of rounding in this case superficially resembles mitotic rounding in involving ERM phosphorylation, retraction fiber formation, and cortical remodeling downstream of ROCK, it does not require Ect2. This work provides insights into the way partial loss of adhesion actives cortical remodeling to drive cell detachment from the substrate. This is important to consider when studying the mechanics of cells in suspension, for example using methods like real-time deformability cytometry (RT-DC).


INTRODUCTION
Once cells enter mitosis, they undergo a rapid change in shape. This occurs throughremodeling of integrin-based substrate adhesions 1 an increase in cortical tension driven by the activation of Myosin II and ERM proteins, 2,3 and an influx of water that leads to an increase in cell volume. 4 Cells in interphase can also be seen rounding up. This can be an active process, e.g. when metastatic cancer cells de-adhere, enabling them to move through the vasculature or lymph. 5 Alternatively, interphase cell rounding can be triggered by changes in the extracellular environment, e.g. following the loss of substrate and/or neighbors. 6,7 Finally, cells also undergo this type of rounding whenever they are removed from the substrate during an experiment, e.g. during cell passaging or when lifting cells into supsension for analysis by real-time deformability cytometry (RT-DC) 8 or cell sorting (FACS).
Although we have a relatively good understanding of the molecular and cellular events that accompany mitotic rounding, many of which occur directly downstream of Cdk1/CyclinB activation, 9 it remains unclear how interphase cells round up. Nevertheless, there are likely to be close parallels between the two processes. Both involve a loss of substrate adhesion. 10 Thus, mitotic cells remain flat when forced to adhere strongly to the substrate. 11 In addition, just as mitotic cells stiffen as they round, [12][13][14] interphase cells have been shown to be less compliant when detached and rounded, than when adhering to a substrate. 15 These changes in cell mechanics likely reflect changes to the actomyosin cortex, 16 whose organization is profoundly affected by mitotic entry, cell shape, external forces, and by both cell-cell and cell-substrate adhesion.
To better understand this process, in this paper we explore the molecular mechanisms that cause interphase cells to round up following their forced removal from the substrate. We show that a partial loss of substrate adhesion triggers actomyosin-dependent cortical remodeling and ERM activation, which facilitates further adhesion loss, as measured by Paxillin-GFP, causing cells to round as they enter suspension. Although the path of rounding in this case superficially resembles mitotic rounding in involving ERM phosphorylation, retraction fiber formation, and cortical remodeling downstream of ROCK, it does not require Ect2. Taken together this work provides insights into the way partial loss of adhesion actives cortical remodeling to drive cell detachment from the substrate.

Modeling interphase rounding
To study the cell shape changes that follow the loss of cell-substrate adhesion, we analysed HeLa cells undergoing detachment induced by various treatments. Although Accutase, EDTA and Trypsin are all effective at inducing de-adhesion and rounding, 17 to image the process live, we searched for a regime in which cell rounding and detachment were robust but occurred with kinetics slow enough to enable live cell imaging (ruling out Trypsin and Accutase at 37 C). This led us to focus our analysis on the effects of Accutase, an enzymatic treatment that induces the digestion of cell-matrix linkers without affecting cell viability or differentiation, 17,18 at 22 C. This is a protocol often used to passage cells and/or when preparing cells for mechanical studies using RT-DC. 8 Importantly, these Accutase-treated cells remained viable for long periods of time in suspension, and were able to respread on transfer onto a substrate ( Figure S1), implying that the treatment does not remove all surface adhesion molecules.

Kinetics of interphase and mitotic rounding
Mitotic rounding was shown previously to require a whole host of changes in cytoskeletal organisation and adhesion to round up. These are triggered downstream of CDK/CyclinB, 9 and brought about via the activation of the RhoGEF, Ect2, 13 and the inactivation of Rap1. 11 To explore whether similar cellular and molecular mechanisms underlie cell rounding in interphase, we began by carrying out live imaging of HeLa cells fluorescently labeled for F-actin following treatment with Accutase at 22 C ( Figures 1A and 1B). This treatment ( Figure 1A) caused cells to undergo a gradual reduction in cross-sectional area ( Figure 1B) and aspect ratio ( Figure 1C) as they rounded up over a period of 15 minutes in a manner that superficially resembled cells entering mitosis.
Although both mitotic cells and cells treated with Accutase left retraction fibers behind as they rounded up, the details of the process differed in the two cases. Thus, although the tips of retraction fibers remained in place in rounding mitotic cells, likely as the result of mitotic-specific changes to integrin-based adhesions 1 , adhesions slipped in cells treated with Accutase ( Figure 1E), as measured by the separation of the tips of retraction fibers from the adhesive interphase cell footprint ( Figures 1D and 1E).
At the same time, we observed an increase in the intensity of actin filaments at the cortex in both Accutasetreated cells and mitotic cells ( Figure 1F), as measured by changes in the localisation of LifeAct-GFP. In addition, like cells entering mitosis, cells treated with Accutase accumulated high levels of the active, phosphorylated form of the cortical membrane-actin linkers ezrin/radixin/moesin (ERM) as they rounded (Figure 1G). Thus, despite the subtle differences in adhesion remodeling in the two cases, cell rounding induced by forced partial de-adhesion of interphase cells from the substrate involves a series of changes to the cortex that resemble those induced upon entry into mitosis.
pERM and ERM kinases LOK/SLK are required for interphase cell rounding ERM proteins, such as Moesin in Drosophila, have previously been shown to be critical for mitotic rounding, 2 whereas phosphomimetic ERM proteins can prevent cells spreading on the substrate. 19 To investigate whether ERM proteins play a similar role in interphase cell detachment, we targeted ERM proteins with the inhibitor NSC668394, which prevents the T567 phosphorylation of ERM family members via direct binding. 20 The addition of NSC668394 to cultures led to a significant impairment of detachment following Accutase treatment ( Figure 2A) in a manner that depended on drug concentration. Thus, when exposed to low concentrations of NSC668394 (10 mM, overnight incubation), cells exhibited slightly reduced detachment dynamics following Accutase treatment, and were more spread than control cells after 20 minutes, the time by which control cells had assumed a spherical shape. Higher concentrations of NSC668394 (250 mM for 3 h) resulted in an even more profound rounding defect ( Figures 2B-D).
Having shown that ERM proteins are activated upon rounding and that their inhibition with NSC668349 delays interphase rounding following Accutase treatment, we next investigated the upstream signaling cascades that might be responsible for ERM activation under these conditions, focusing on LOK (serine threonine kinase 10 STK10, lymphocyte oriented kinase LOK) and SLK (STE20-like kinase) kinases, which have been shown to phosphorylate ERM proteins in many systems. 21,22 To silence the expression of both LOK and SLK, we combined siRNAs against the two kinases, and used siRNA against GAPDH as a control. The efficacy of the LOK/SLK siRNA effect was confirmed by western blotting. RNA-mediated gene silencing led to a significant loss of SLK protein, partial decrease in LOK protein, and was associated with a corresponding decrease in pERM levels ( Figure 2I). This silencing of LOK and SLK was also found to impair both mitotic (FiguresS2A and S2B) and interphase rounding following a 4 minute treatment with Accutase ( Figures 2E-2H). After 15 minutes, however, LOK/SLK siRNA treated cells had assumed a spherical shape, resembling that of GAPDH siRNA control cells ( Figure 2J). This iScience Article indicates that, once detached from the substrate, pERM proteins are not required for interphase cells to assume a completely spherical shape (although these cells appeared slightly smaller than controls ( Figure  2K)). This was the case even though LOK/SLK siRNA cells in suspension had a markedly reduced ratio of cortical to cytoplasmic pERM ( Figure 2L) and diminished levels of cortical pERM ( Figure 2M). Taken together these data suggest that pERM levels become elevated as cells round in a manner that depends on LOK/SLK, but which is independent of the trigger (partial disassembly of adhesions or entry into mitosis 23 ). Mitotic rounding also depends on actin and ROCK-mediated contractility. 24 To test if the same is true during the interphase rounding induced by partial de-adhesion from the substrate, we pre-treated HeLa cells with the ROCK inhibitor Y27632 to decrease myosin activity, or with the actin monomer sequestering agent Latrunculin B (Lat-B) to depolymerize the actin cytoskeleton ( Figure 3A). Twenty minutes later, cells were then treated with Accutase and imaged as they detached from the substrate (in the constant presence of inhibitors) ( Figure 3B). In this experiment, the inhibition of ROCK using Y27632 (which decreased levels of cortical Myosin II, Figure 3C) led to a significant impairment in cell rounding relative to the corresponding control. As a result, Accutase-treated cells remained attached to spread on the substrate ( Figures 3B  and 3C). Disruption of the actin cytoskeleton using Lat-B reduced levels of both cortical actin and Myosin II and had a similar effect ( Figures 3B and 3C). To better understand the impact of ROCK on cell rounding, we next measured cell shape changes in fixed Y27632-treated or DMSO-treated (control) cells exposed to Accutase for 4 minutes (Figures 3D-3F). Y27632-treated cells had a higher surface area ( Figure 3D), a higher aspect ratio ( Figure 3E), and a decreased cell height ( Figure 3F). These data highlight the roles of both ROCK and actin in the process of interphase cell rounding up during detachment from the substrate.
Because both pERM proteins and active Myosin II have been reported to alter cell-substrate adhesion, it was important to determine whether or not Y27632-treated cells and NSC668349-treated cells remain flat and attached to the substrate after Accutase treatment as the result of increased interphase adhesion. To this end, we imaged Paxillin dynamics in control interphase cells and in cells treated with the two inhibitors before and after the addition of Accutase ( Figure S3). Adhesions appeared similar in control and NSC668349-treated cells, but were markedly reduced in cells treated with Y27632, as shown previously. 25 Thus, there was no correlation between the rate of rounding and the size of interphase adhesions as measured by Paxillin-GFP (Figure S3). In addition, adhesions in these experiments were seen disassembling as cells underwent rounding, but did not change in cells that remained flat. This implies that the cortical forces inducing rounding are important for adhesion disassembly in cells treated with Accutase. In line with this, when we induced the slow detachment of cells from the substrate using the calcium chelator EDTA instead of Accutase (Figur-esS4A-S4E), we observed much slower loss of Paxillin-GFP adhesions. In this case, too, adhesion remodeling and rounding depended on ROCK. These data suggest that there is positive feedback in the system, as the loss of adhesions triggers cortical contractility, which further reduces cell-substrate adhesion. Note that although adhesions in Y27632 treated cells following the addition of Accutase were still present, they appeared visibly smaller than those in control cells. Thus, although cells remained flat and attached to the substrate long after Accutase addition, they could be detached using mechanical force (banging the plate).
To compare the cell biology of DMSO-treated and Y27632-treated cells detached using this shake-off method in detail, we incubated cells in suspension for different times (immediately, t = 0 min, or at t = 30 min), before fixing and staining them to monitor cell shape and cytoskeletal organization. The confocal microscopy analysis of these cells revealed that although DMSO-treated cells round up as they detach, it takes time for Y27632-treated cells to fully round following detachment ( Figures S5A and S5B).
After remaining in suspension for 30 min, however, both DMSO and Y27632 cells appear as near perfect spheres.
When we monitored the localization of the active phosphorylated ERM proteins in suspension HeLa cells +/À ROCK inhibitor, we observed relatively uniform and elevated phosphorylated ERM staining in DMSO-treated cells at early timepoints after the Accutase treatment. By contrast, the cortical recruitment of pERM proteins appeared incomplete in Y27632-pre-treated cells immediately after detachment from the substrate  iScience Article ( Figure S5D). As they rounded, however, these cells accumulated relatively uniform and high levels of cortical pERM (FiguresS5C and S5D). Furthermore, the increase in cortical accumulation of pERM in these ROCK inhibitor-treated cells was found to correlate with cell shape as measured by cell aspect ratio ( Figure S5E). These data imply that ERM activation is responsive to cell shape (as has been suggested previously 26 ), and does not strictly require the activity of ROCK. Thus, although ROCK is required for cells to round following the loosening of their adhesions, and accelerates de-adhesion and cell rounding in suspension, ROCK is not required for detached cells to assume a final spherical shape -which represents a minimal energy conformation.

Ect2 is not required for interphase cell rounding
The data presented thus far show that interphase cell rounding following Accutase treatment is driven by similar processes as mitotic rounding. A common trigger is the weakening of substrate adhesions, which occurs either through enzymatic digestion (Accutase), or by adhesion remodeling downstream of Cdk1/ CyclinB activation (mitosis). In addition, rounding in both cases is promoted by the activation of the ROCK-Myosin II and SLK/LOK-dependent pERM activation. These similarities are surprising given the differences in the mechanical properties of the cortex in rounded interphase cells and mitotic cells. 27 Given this, we thought it important to identify differences in the two modes of rounding. As shown previously, Ect2 knockdown impacts the rounding of mitotic cells. 13 In our interphase experiments, Ect2 knockdown ( Figures 4B and 4C) had no visible impact on the detachment of HeLa cells from surface following Accutase treatment ( Figure 4A). Interphase Ect2 RNAi cells were of similar size and shape to control cells at the beginning of Accutase treatment, and decreased their surface area at the same rate as control cells while rounding up during detachment (Figures 4D and 4E). This is in line with the idea that Ect2 is largely inactive in interphase, where it is localized largely to nucleoli, 28 and is first activated in the cytoplasm at G2/M. These data suggest that different cues can trigger the same type of cortical remodeling to drive cell rounding.
Finally, to determine if the phosphorylation-induced activation of Myosin II and ERM are sufficient to induce rounding, we performed the time-course experiments following the addition of Calyculin A to cell cultures ( Figure 3G). Calyculin A is known to increase levels of activated Myosin, 29,30 pERM 19 and to reduce integrin-mediated cell adhesion. 19 As observed previously with 3T3 fibroblasts 19,29 and HEK293T iScience Article cells, 19 increasing cell contractility (via treatment with Calyculin A) was found to be enough to drive detachment, causing adherent HeLa cells to round up ( Figures 3G and 3H).

Conclusions
Taken together, these data suggest that cells entering mitosis and interphase cells rounding following deadhesion follow a strikingly similar path of cell shape remodeling. In both cases, cells leave retraction fibers behind as they round. Moreover, rounding in the two cases is driven by a common process of actomyosin remodeling. In both cases, the trigger to round induces a similar upstream pathway (ROCK and LOK/SLK) leading to the activation of Myosin and pERM proteins respectively. In our data, the accumulation of pERM at the cortex was correlated with cell shape as measured with aspect ratio, and so may be a consequence as well as a cause of the rounding, providing positive feedback to the system. We speculate that the assembly of a rigid round cortex that is rich in actin and ERM proteins in both cases may be part of a general mechanism cells use to ensure that they are able to resist external forces and survive under crowded conditions (mitosis) and when subject to shear and flow (suspension, circulating metastatic cells).
The similarities between mitotic and interphase rounding are surprising given the differences in the reported cell biology of round interphase and mitotic cells, which include differences in actin cortex thickness, 3 composition, 31,32 filament organisation, 3,27 stiffness, 12-14 pressure 33 and cytoplasmic viscosity. 34 However, we also observed clear differences in specific aspects of the process of rounding in the two cases. First, the tips of the retraction fibers formed in Accutase treated cells were found to slide as cells rounded. By contrast, the tips of retraction fibers largely remained in place in cells entering mitosis, likely because of a specific program of mitotic cell adhesion remodeling, 1 which is important to provide cells with a memory of the mother cell footprint as they divide. Second, interphase rounding induced by detachment from the substrate does not require Ect2 as an upstream trigger for ROCK activation. This makes sense, given the need for CDK1/ CyclinB to trigger the nuclear export of Ect2 protein so that it can activate Rho and ROCK in the cytoplasm. This raises the question of nature of the trigger inducing activation of ERM proteins and Myosin II in interphase cells following their partial detachment from the substrate. Given the marked differences in the rates of actomyosin-dependent rounding triggered by Accutase and EDTA, we speculate that this may be a direct effect of the weakening of adhesions on signaling that impacts the actomyosin cortex. The disassembly of adhesions leads to activation of Myosin II, which in turn leads to a mechanically induced further loss of adhesions, and to rounding, which activates ERM proteins. Although a similar cascade may be triggered at mitotic entry, cells counter it by remodeling adhesions to ensure that some remain in place to guide respreading. In this way, our study helps to highlight the commonalities and important regulatory differences between these two types of rounding.
We note that although this analysis of interphase rounding involves studying the effects of an artificial treatment to induce the partial detachment of cells from the substrate, our findings are important to consider when interpreting experiments that involve the analysis of cells placed in suspension (e.g. RT-DC) -which as we show here have undergone profound and time-dependent changes in their cortical organization as they round both on the substrate and once they are in suspension. In addition, this work is likely to be relevant to the study of cancer cells which switch between adhesion-dependent and independent modes of motility as they undergo metastatic spread.

Limitations of the study
There are several important limitations to our study. First, this study focused on the rounding following detachment in HeLa cells -a single cancer-derived cell line. The mechanisms described here might be different in normal epithelial cells or cells of different origin (e.g., mesenchymal). Second, by applying Accutase to cells at 22 C, we induced a relative slow cell detachment through partial de-adhesion from the substrate to better study the process in-depth without the induction of cell blebbing. Some of the observations made may depend on the precise kinetics of rounding. Third, in the future it will be important to test the relevance of the mechanisms of interphase cell detachment described here in a more physiological 3D environment, e.g., in organoids, spheroids or in vivo models.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

Lead contact
Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Buzz Baum (bbaum@mrc-lmb.cam.ac.uk).

Materials availability
This study did not generate new unique reagents.
Data and code availability d Microscopy data and western blot images reported in this paper will be shared by the lead contact upon request.
d The paper does not report original code.
d Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Perturbation experiments
To inhibit ROCK, cells were pre-treated for 20 min with Y27362 (Cat#Y0503, Sigma-Aldrich) at a final concentration of 10 mM. For suspension analysis, following detachment with Accutase (Cat#A1110501, ThermoFisher Scientific) for 5 min, cells were collected into complete DMEM (+/À Y27362), centrifuged at 500 xg for 5 min and resuspended in complete DMEM (+/À Y27362). To increase cell contractility, Calyculin A (Cat#C5552, Sigma-Aldrich) at final concentration of 25 nM was used and live imaging started immediately following its addition. Ezrin was inhibited with NSC668394 (Cat#341216, Merck Millipore) overnight at a final concentration of 10 mM or for 3 hr at 250 mM. For actin depolymerisation Latrunculin B (Cat#L5288, Sigma-Aldrich) was used at a final concentration of 20 nM. Equivalent volume of DMSO was used as controls. Accutase treatment and following incubation in suspension were performed at room temperature (22 C-25 C, RT). Accutase treatment had no effect on cell viability or ability to respread ( Figure S1). Cells were also detached using EDTA (0.68M, in-house prepared solution).