A flexible network of Vimentin intermediate filaments promotes the migration of amoeboid cancer cells through confined environments

The spread of tumor cells to distant sites is promoted by their ability to switch between mesenchymal and amoeboid (bleb-based) migration. Because of this, inhibitors of metastasis must account for each motility mode. To this end, here we determine the precise role of the Vimentin intermediate filament system in regulating the migration of amoeboid human cancer cells. Vimentin is a classic marker of epithelial to mesenchymal transition and is therefore, an ideal target for a metastasis inhibitor. However, the role of Vimentin in amoeboid migration has not been determined. Since amoeboid, leader bleb-based migration occurs in confined spaces and Vimentin is known to be a major determinant of cell mechanical properties, we hypothesized that a flexible Vimentin network is required for fast amoeboid migration. This was tested using our PDMS slab-based approach for the confinement of cells, RNAi, over-expression, pharmacological treatments, and measurements of cell stiffness. In contrast to Vimentin RNAi, inducing the bundling of Vimentin was found to inhibit fast amoeboid migration and proliferation. Importantly, these effects were independent of changes in actomyosin contractility. Collectively, our data supports a model whereby the perturbation of cell mechanical properties by Vimentin bundling inhibits the invasive properties of cancer cells.


Introduction
Cell migration is required for embryonic development, immune surveillance, and wound healing in healthy individuals. However, the uncontrolled migration of tumor cells to distant sites is a hallmark of metastasis and is associated with poor prognosis. In recent years, it has been demonstrated that cancer cells can switch between focal adhesion (mesenchymal) and blebbased (amoeboid) migration modes. This is important because blocking metastasis will require that each mode of migration be targeted. To this aim, here we determine the role of a wellestablished regulator of mesenchymal migration, the Vimentin Intermediate Filament (VIF) cytoskeleton, in regulating the amoeboid migration of cancer cells.
The switch from a predominantly Keratin to Vimentin expression pattern is a classic marker of Epithelial to Mesenchymal Transition (EMT). Accordingly, Vimentin is known to increase the size and strength of focal adhesions 23 . In contrast, the role of Vimentin in amoeboid (blebbing) cells has not been determined. Using in vivo and in vitro approaches, it has been shown that highly contractile (metastatic) cancer cells will switch from a mesenchymal to "fast amoeboid" mode of migration in response to physically confining environments, such as those found in micro-lymphatics/capillaries and perivascular spaces 2,17,19,24,26 . Additionally, certain drug treatments including, Matrix Metalloprotease (MMP) and tyrosine kinase inhibitors (e.g., Dasatinib), will induce a switch to bleb-based migration 6,20,29 . Fast amoeboid migration relies on the formation of what we termed a leader bleb 19 . In confined environments, leader blebs are typically very large and stable blebs containing a rapid cortical actomyosin flow 2,17,19,26 . Whereas mesenchymal cells utilize integrin-Extracellular Matrix (ECM) interactions for migration, fast amoeboid or Leader Bleb-Based Migration (LBBM) only requires friction between the cortical actomyosin flow and the extracellular environment 2 . This property likely promotes the invasive properties of cancer cells in vivo.
Because metastasis requires that cells migrate within the confines of tissues, we hypothesized that confined cancer cell migration (i.e., LBBM) requires a flexible intermediate filament network. Unlike Keratin, which stiffens by bundling in response to force (i.e., strain stiffens), Vimentin remains unbundled and flexible 21 . Moreover, photobleaching experiments have shown that Vimentin undergoes subunit exchange an order of magnitude faster than Keratin and are therefore, considered to be more dynamic 16 . Recently, a statin used for lowering blood cholesterol, Simvastatin, was identified in a screen for Vimentin binding molecules 28 . In contrast to other statins, such as Pravastatin, Simvastatin was found to bind the sides of Vimentin and induce bundling. In cell-based assays, Simvastatin was shown to block the proliferation of adrenal carcinoma cells, possibly because Vimentin bundling inhibits its degradation required for cell division 5 . Importantly, Simvastatin binds Vimentin with high specificity, as opposed to other molecules (e.g., Withaferin A) that effect other components of the cytoskeleton 3,13 . Here, by combining Simvastatin with our recently described approach for the confinement of cells, we describe the precise role of a flexible (unbundled) Vimentin network in amoeboid human cancer cells 18 .
Our data show that the concentration of Vimentin and its bundling are potent regulators of mesenchymal and amoeboid migration, mechanics, and the survival of human cancer cells in confinement. Collectively, this work sheds new light on the potential of Vimentin as a therapeutic target.

Results and Discussion
Because a high level of Vimentin expression is correlated with hematogenous metastasis within a wide array of melanoma samples, we set out to determine the localization of Vimentin in melanoma A375-M2 cells 15 . Moreover, this highly metastatic sub-line has been observed by intravital imaging to undergo amoeboid migration in tumors 27 . Using Vimentin tagged on its C-terminus with FusionRed, Vimentin-FusionRed, and transient transfection we determined the localization of Vimentin in A375-M2 cells by live high-resolution microscopy ( Fig.   1). In cells adhered to fibronectin coated glass, an isotropic network of Vimentin was concentrated near the cell center (Fig. 1A, left). Similarly, in non-adherent (blebbing) cells on uncoated glass, Vimentin surrounded the nucleus and was excluded from blebs ( Fig. 1A, middle). In order to evaluate the localization of Vimentin in cells with leader blebs, we promoted the conversion of A375-M2 cells to this morphology by confinement using our Polydimethylsiloxane (PDMS) slab-based approach 18 . This involves placing cells between a Bovine Serum Albumin (BSA; 1%) coated slab of PDMS and cover glass, which is held at a defined height by beads with a diameter of ~3 µm. This confinement height was shown to be optimal for stimulating the transition to fast amoeboid migration 17 . Using this approach, we found that Vimentin was kept entirely within the cell body as opposed to leader blebs (Fig. 1A, right & Movie S1). This is significant because the cell body is the principle source of resistance to the cortical actomyosin flow in leader blebs, limiting LBBM speed 17 .
To directly test the notion that Vimentin concentration in the cell body limits LBBM speed, we depleted A375-M2 cells of Vimentin using a Locked Nucleic Acid (LNA), which offer enhanced specificity and stability over traditional small interfering RNAs (siRNAs) 10 . Because of the long half-life of Vimentin, cells were incubated with LNAs for 5 days to achieve a ~90% reduction in protein levels (Fig. 2B). Moreover, because these cells predominantly express Vimentin, they are an ideal (simplified) model for defining the role of intermediate filaments in LBBM (Fig. 1B). Using our PDMS slab-based approach, LBBM was quantitatively evaluated for LNA treated cells by live imaging over 5 hr. Strikingly, in cells depleted of Vimentin, the speed of LBBM was increased over control by ~50% (Fig. 2E-F & Movie S2), whereas directionality remained the same for each group (Fig. S1A). Quantitation proved that leader bleb area, which is defined as the single largest bleb within a given frame, is close to double the size of control ( Fig. 2C). Interestingly, quantitation of cell body area found that in Vimentin RNAi cells, the cell body area was decreased by over 25% (Fig. 2D). This result is consistent with the location of Vimentin in these cells, which may limit the degree to which cortical actomyosin is able to contract the cell body. Consequently, more cytoplasm from the cell body can enter leader blebs, increasing their size. Strikingly, the nucleus in Vimentin RNAi cells was observed to undergo large shape changes, which may reflect an increase in the degree of force transmitted to the nucleus from cortical actomyosin ( Fig. 2A). A more than 25% increase in the number of Vimentin RNAi cells undergoing apoptosis is consistent with reports of nuclear rupture and DNA damage in confined cells ( Fig. S2A) 8,25 . To test the hypothesis that Vimentin regulates the stiffness of A375-M2 cells, we used an approach described by the Piel Lab (Institut Curie) that involves compressing cells between two Polyacrylamide (PA; 1 kPa) gels ( Fig. 2G) 17 . Using this approach, cell height divided by the diameter (h/d), which is a function of the opposing force, is used to define the "cell stiffness." Consistent with other reports, we find that cells depleted of Vimentin are ~25% softer than control cells (Fig. 2H). Because cortical actomyosin is also expected to regulate stiffness, we confirmed that the level of active (phosphorylated) Regulatory Light Chain (p-RLC) is not affected by Vimentin RNAi (Fig. 2I). Therefore, Vimentin expression increases cell stiffness to limit migration in confined environments. Simvastatin were ~60% slower than those treated with Vehicle and Pravastatin (Fig. 4D). To evaluate the effect of Simvastatin on cell stiffness, we again used the gel sandwich approach.
For cells treated with Simvastatin, we observed a small decrease (~10%) in cell stiffness, whereas treating cells with Pravastatin had no effect when compared to Vehicle (Fig. 4E).
Because the bundling of Vimentin causes the network to collapse into a small area of the cytoplasm, this result might be expected since the gel sandwich assay measures the stiffness of the entire cell. However, since bundled intermediate filament networks are known to be much stiffer, a local (large) increase in mechanical properties is expected 21 . In line with this concept, we observe leader blebs to pull against sites of Vimentin bundling in the cell body (Fig. 4B &   Movie 4). Because the actomyosin cytoskeleton is also expected to effect cell stiffness, we also measured the level of active myosin (p-RLC) in drug treated cells. By Western blotting, we confirmed that treatment with Vehicle, Simvastatin, and Pravastatin did not affect the level of p-RLC (Fig. 4F). Moreover, the concentration of Vimentin was unchanged in these cells (Fig. 4F).
Therefore, by inducing Vimentin bundling with Simvastatin, migration in confined environments is inhibited. As tissue culture cells obtain cholesterol from serum, statin treatment is not expected to alter the level of intracellular cholesterol, which is a critical component of the plasma membrane 7 . Therefore, our data supports a model whereby local stiffening of the cell body by Vimentin bundling inhibits LBBM (Fig. 5E).
In order to evaluate if Simvastatin has a general effect on cancer cell motility, we subjected drug treated A375-M2 and lung cancer A549 cells, which also express Keratin (Fig.   1B), to transmigration assays. Using this approach, we found that Simvastatin decreases transmigration for A375-M2 (~20% of control) and to a lesser extent A549 cells (~85% of control), whereas Pravastatin did not have a significant effect (Fig. 4A-B). Because we observed a ~50% increase in the number of apoptotic cells after Simvastatin treatment (Fig.   S2C), we next determined if Simvastatin had a general effect on cell proliferation. To accomplish this, we counted cells over 5 consecutive days in order to generate growth curves for A375-M2 and A549 cells. Strikingly, we found that Simvastatin but not Pravastatin treatment inhibited the proliferation of both cell types (Fig. 5C-D). However, this effect was significantly more pronounced for A375-M2 cells, which predominantly express Vimentin (Fig. 1B).
Consistent with Vimentin increasing the size and strength of focal adhesions, A375-M2 and A549 cells were frequently de-adhered one day after Simvastatin treatment (Fig. 5E) 23 . Again, this effect was much more pronounced for A375-M2 cells. In contrast, cells treated with Pravastatin were not significantly different from Vehicle treated (Fig. 5E). Altogether, these results demonstrate that the migration and proliferation of cancer cells is inhibited by Vimentin bundling.
To the best of our knowledge, this study is the first to describe Vimentin bundling as a

Leader bleb, cell body, and Vimentin area measurements
For leader bleb, cell body, and Vimentin areas, freshly confined cells were traced from highresolution images with the free-hand circle tool in Fiji (https://fiji.sc/). From every other frame, the percent of cell body area for leader blebs, percent of total for cell body areas, and percent of cell body area for Vimentin was calculated in Microsoft Excel (Redmond, WA). Frame-by-frame measurements were then used to generate an average for each cell. All statistical analyses were performed in GraphPad Prism (La Jolla, CA).

Cell migration
To perform cell speed and directionality analyses, we used a Microsoft Excel plugin, DiPer, developed by Gorelik and colleagues and the Fiji plugin, MTrackJ, developed by Erik Meijering for manual tracking 12,22 . For minimizing positional error, cells were tracked every other frame.
Brightfield imaging was used to confirm that beads were not obstructing the path of a cell. All statistical analyses were performed in GraphPad Prism.

Cell stiffness measurements
The gel sandwich assay has been described in detail elsewhere 17 .

Transmigration
Prior to transmigration assays, polycarbonate filters with 8 µM pores (cat no. 83-3428; Corning, Corning, NY) were coated with 10 µg/mL fibronectin (Millipore) by air drying for 1 hr. After permitting ~100,000 cells in serum free media to attach (1 hr), DMSO, Simvastatin, or Pravastatin were added to each well. Bottom chambers contained 20% FBS in media to attract cells. After 24 hr, A375-M2 or A549 cells from the bottom of the filter were trypsinized and counted using an automated cell counter (TC20; Bio-Rad, Hercules, CA). Transmigration was then calculated as the ratio of cells on the bottom of the filter vs. the total. All statistical analyses were performed in GraphPad Prism.

Growth curves
On day zero, ~125,000 cells were plated in 6-well tissue culture plates in complete media with DMSO, Simvastatin, or Pravastatin. For 5 consecutive days, A375-M2 or A549 cells were trypsinized and counted using an automated cell counter (TC20; Bio-Rad). Each day, wells were supplemented with fresh media and drug till their day to be counted. All plots were generated using GraphPad Prism.

Statistics
All sample sizes were empirically determined based on saturation. Outliers were identified by the ROUT method in GraphPad Prism and excluded from further analyses. As noted in each figure legend, statistical significance was determined by either a two-tailed (unpaired) Student's t-test, F test, or ordinary one-way ANOVA followed by a post-hoc multiple comparisons test.