Vimentin filaments drive migratory persistence in polyploidal cancer cells

Significance

Polyploidal cells have been shown to correlate with poor patient outcomes in the clinic, as well as with increased chemoresistance and tumorigenicity. They are also implicated in cancer aggression due to their highly migratory phenotype. There is a serious lack of understanding of the underlying biophysical mechanisms that drive their increased migration. Here we address this lack of knowledge and show that vimentin is key to their migratory phenotype, especially in the context of cellular jamming. We also propose a therapeutic avenue that can be adapted in the clinic to mitigate the deleterious effects of these giant cancer cells on patient outcomes.

Abstract

Polyploidal giant cancer cells (PGCCs) are multinucleated chemoresistant cancer cells found in heterogeneous solid tumors. Due in part to their apparent dormancy, the effect of PGCCs on cancer progression has remained largely unstudied. Recent studies have highlighted the critical role of PGCCs as aggressive and chemoresistant cancer cells, as well as their ability to undergo amitotic budding to escape dormancy. Our recent study demonstrated the unique biophysical properties of PGCCs, as well as their unusual migratory persistence. Here we unveil the critical function of vimentin intermediate filaments (VIFs) in maintaining the structural integrity of PGCCs and enhancing their migratory persistence. We performed in-depth single-cell analysis to examine the distribution of VIFs and their role in migratory persistence. We found that PGCCs rely heavily on their uniquely distributed and polarized VIF network to enhance their transition from a jammed to an unjammed state to allow for directional migration. Both the inhibition of VIFs with acrylamide and small interfering RNA knockdown of vimentin significantly decreased PGCC migration and resulted in a loss of PGCC volume. Because PGCCs rely on their VIF network to direct migration and to maintain their enlarged morphology, targeting vimentin or vimentin cross-linking proteins could provide a therapeutic approach to mitigate the impact of these chemoresistant cells in cancer progression and to improve patient outcomes with chemotherapy.

Intratumor heterogeneity is one of the greatest obstacles facing cancer therapy. This heterogeneity in cancer cells stems from the clonal expansion of genetically diverse cells under unique environmental pressures (1–3). These mosaic tumors often include subpopulations of cancer cells that adapt to harsh conditions by developing aggressive and chemoresistant phenotypes. The tumor microenvironment is also heterogeneous in structure and composition (4). This leads to complicated molecular landscapes within the tumor, with subpopulations of tumor cells that are highly resistant to chemotherapy.

One such subpopulation of drug-resistant cancer cells comprises polyploidal giant cancer cells (PGCCs). This small subpopulation of multinucleated cancer cells is ubiquitous in solid tumors (5–10). Recent publications have highlight the chemoresistant and tumorigenic nature of these giant cells (10, 11). PGCCs that remain dormant in tissues can undergo amitotic budding to give rise to new tumors (8) with increased genetic diversity (12), which further contributes to the heterogeneity in solid tumors. Despite recent advancements, there remains a significant lack of understanding of the role these giant cells play in cancer progression and metastasis.

In a previous paper, we reported that PGCCs exhibit a unique cytoskeletal organization, giving rise to elevated stiffness and migratory persistence (10). These findings prompted us to further examine their unique biophysical properties and migratory phenotype, as these traits are closely linked with cancer invasion and metastasis (13, 14). Specifically, we wanted to look at vimentin intermediate filaments (VIFs), which play a critical role in cancer cell motility and metastasis (15, 16). Previous studies have demonstrated that VIFs orchestrate microtubule patterning and alignment of traction stresses during cell migration (17). Jamney et al. (18) showed decreased nuclear and cellular volume in murine embryonic fibroblasts from vimentin knockout mice, suggesting an important role for vimentin in regulating cell structure.

Vimentin has unique strain-hardening properties that allow VIFs to act as shock absorbers in the cell, protecting the cell against compressive loads and increasing cytoplasmic viscosity under mechanical stress (19, 20). As such, the presence of VIFs is crucial for maintaining overall cell shape and integrity. Based on the previous vimentin studies, we postulated that PGCCs may have unique adaptations in their vimentin structure to help support their enlarged morphology and drive their increased migration. To test this hypothesis, we performed an in-depth single-cell analysis to examine the distribution of VIFs in PGCCs isolated from the triple-negative breast cancer cell line MDA-MB-231. We found that PGCCs have significantly increased levels of cytoplasmic vimentin, as well as a more diffuse, evenly distributed organization of VIFs compared with non-PGCC controls. Furthermore, these enlarged cells have a more polarized VIF network during migration, and disruption of VIFs leads to reduced cell volume and loss in directed cell migration. Our results provide insight into the role of VIFs in driving PGCC behavior and shed light on potential therapeutic avenues that can be used to mitigate the role of these chemoresistant cells in cancer progression (Fig. 1).

Fig. 1.

Schematic depicting migratory persistence and unique vimentin distribution in PGCCs. Compared with non-PGCCs, PGCCs migrate more consistently into the scratch wound and have higher expression and more distributed vimentin.

Results

PGCCs Express and Maintain a Highly Chemoresistant Phenotype with Mixed Epithelial-to-Mesenchymal Transition Signatures.

To understand the chemoresistant nature of PGCCS and their latent role in cancer relapse, a purified population of MDA-MB-231 PGCCs was generated via treatment with high-concentration (500 nM) paclitaxel (PTX) for 18 h, followed by a 7-d recovery period. For the purposes of this study, PGCCs were defined as cells containing >2.5 times the DNA quantity of the average population, as determined with DAPI or Hoechst staining, consistent with our previous report (10). The fraction of PGCCs increased steadily after treatment, and the population remained stably pure up to 1 wk. The PGCC percentage then began to drop after approximately 14 d, due to amitotic budding and generation of daughter cells (11). By week 8 after initial treatment, relatively low amounts of PGCCs remained, signaling a return to equilibrium (Fig. 2A).

Fig. 2.

Quantification of PGCC formation and recovery after treatment, which express a mixed-EMT profile. (A) PGCC recovery after initial treatment with 500 nM PTX quantified from DAPI- stained MDA-MB-231 nuclei. (B) MTT analysis of cell viability of PGCC at 7 d after treatment with 500 nM PTX and of untreated non-PGCCs. (C) qRT-PCR analysis of MDA-MB-231 PGCCs relative to non-PGCCs, normalized to respective 18S rRNA controls. All error bars represent SEM. (D and E) MTT analysis of cell viability of controls and PGCCs in MDA-MB-453 (D) and SKBR (E) cells. **P < 0.01; ***P < 0.001, Student’s t test.

To assay the susceptibility of our purified PGCC population to PTX, we performed a standard MTT assay at various PTX concentrations and measured their viability after 48 h. We found that while nonpolyploidal MDA-MB-231 cells (non-PGCCs) experienced significant levels of cell death, PGCCs maintained almost 100% viability even at the highest PTX concentration (100 nM) (Fig. 2B). This finding was corroborated by studies performed in HER-2⁺ SKBR-3 breast cancer cells (Fig. 2 D and E), further highlighting the ubiquitous chemoresistant nature of PGCCs.

Next, we looked at the expression of genes associated with the epithelial-to-mesenchymal transition (EMT), which is a critical step in cancer progression to metastasis (21–23). To determine the expression levels of various EMT genes, we performed qRT-PCR analysis in PGCCs and non-PGCC control cells. PGCCs presented a mixed-EMT pattern, with up-regulation of mesenchymal genes fibronectin, N-cadherin, and vimentin; up-regulation of epithelial marker E-cadherin; and reduced FAK expression (24, 25) (Fig. 2C).

PGCCs Migrate Preferentially into the Scratch Wound Even after PTX Treatment.

Our previous study used the persistent random walk model to analyze cell tracking data to show that PGCCs undergo more persistent migration than non-PGCCs. To further examine this migratory persistence and to determine whether it is affected by PTX treatment, we performed a 12-h scratch wound migration assay with periodic imaging at 10-min intervals. Under control conditions, MDA-MB-231 cells were able to cover a significant portion of the scratch wound in 12 h (Fig. 2 A and D and Movie S1); however, when cells were treated with 25 nM PTX, we observed a small reduction in coverage area from 9 to 12 h (Fig. 2 B and D), indicating that PTX had inhibited migration.

Using single-cell analysis via CellProfiler (Broad Institute), we were able to characterize the population of cells that had migrated into the scratch wound along with their nuclear area (Fig. 2C). The migrated cell population was sorted based on this nuclear area to identify the subpopulations of PGCCs and non-PGCCs. This quantification revealed a significant increase in the percentage of PGCCs and a corresponding decrease in the percentage of non-PGCCs that had migrated into the scratch during the 12-h period (Fig. 3 E and F). With 25 nM PTX treatment, the percentage of PGCCs that had migrated into the scratch wound was further increased by 10%, suggesting that their migration was not affected by PTX treatment (Fig. 3 E and F). Taken together, these findings demonstrate the PTX-resistant and highly migratory phenotype of PGCCs.

Fig. 3.

PGCCs preferentially migrate into scratch wound. (A and B) DAPI-stained fluorescent images of control (A) and 25 nM PTX-treated MDA-MB-231 (B) cells at 12 h after scratch in a wound healing assay. (C) Representative CellProfiler analysis of identified nuclei, cells, coverage area, and migrated cells during the assay. (D–F) Quantification of coverage area (D), fraction of PGCCs in migrated cells (E), and non-PGCC fraction (F) every 3 h after the initial scratch. n ≥ 500 cells for all conditions. All error bars represent SEM. **P < 0.01; ***P < 0.001, Student’s t test.

PGCCs Exhibit Unique Vimentin Organization and Increased Cell and Vimentin Volume.

We next investigated the underlying mechanisms driving the increased directional migration and unusual biophysical phenotype of PGCCs. Our PCR analysis showed increased vimentin expression in PGCCs relative to non-PGCCs (Fig. 2C). We hypothesized that PGCCs rely on a uniquely distributed network of VIFs to maintain their increased volume and highly migratory phenotype. To study the organization of VIFs, both PGCC and non-PGCC populations were fixed and stained with a vimentin antibody before imaging with epifluorescent microscopy (Fig. 4 A–D). The vimentin structure was highly condensed near the perinuclear region in non-PGCCs, whereas vimentin spread throughout the cytoplasm of PGCCs. Quantification of the fluorescence signal based on integrated area per cell also showed increased vimentin expression in PGCCs compared with non-PGCCs (Fig. 4E); however, the mean vimentin intensity, which characterizes vimentin expression per pixel, was similar in PGCCs and non-PGCCs (Fig. 4F). This indicates that although the overall vimentin expression is higher in PGCCs, the VIF network is more tightly clustered in non-PGCCs. Within regions occupied by this VIF network, the average intensity of vimentin polymers was lower in PGCCs than in non-PGCCs, in part to the more diffuse structure of the VIF network in PGCCs. Confocal imaging (Fig. 4G) allowed us to create 3D volume reconstructions to quantify cell volume (Fig. 4H), in which we observed significantly larger volumes for PGCCs compared with non-PGCCs (Fig. 4I). Finally, 3D volume analysis of the VIF network corroborated our 2D findings showing that VIFs occupied a significantly larger portion of the cell volume in PGCCs compared with non-PGCC controls (Fig. 4J), even when normalized to overall cell volume.

Fig. 4.

PGCCs express increased but more distributed vimentin compared with non-PGCCs. (A–D) Confocal microscopy images of non-PGCCs (A and B) and PGCC MDA-MB-231 cells (C and D) stained for DNA (DAPI; cyan), actin (rhodamine phalloidin; red), and vimentin (green). (E and F) Quantification of total vimentin intensity (n > 50 cells) (E) and mean intensity of nonpolyploidal and polyploidal cells (n > 50 cells) (F). (G) Representative confocal Z-stack image of MDA-MB-231 cells stained for DNA (DAPI; cyan), actin (rhodamine phalloidin; red), and vimentin (green). (H) ImageJ 3D volume reconstruction of MDA-MB-231 cells based on confocal images of actin-stained cells. (I and J) Quantification of total cell volume (I) and vimentin volume (J) in non-PGCCs (n = 36 cells) and PGCCs (n = 32 cells). All error bars represent SEM. ns, not significant; **P < 0.01; ***P < 0.001, Student’s t test.

Altered Distribution and Increased Polarization of Vimentin in Migrating PGCCs.

To further investigate the role of VIFs in PGCC migration, we fixed and stained MDA-MB-231 cells migrating in the scratch wound at different time points with vimentin antibody. Recent studies have found that the transition from static (jammed) to migratory (unjammed) states involves alterations in cytoskeletal organization (26); thus, we examined the cytoskeletal features of cells in both the jammed and unjammed states, as well as after their migration into the scratch wound, to understand the role of VIFs in PGCC and non-PGCC migration.

We divided our scratch wound assay into three primary regions. The “jammed” region consists mainly of cells located far from the edge of the scratch wound, in regions of the monolayer with high cell packing density and more constrained motion. In the “unjammed” region at or near the scratch wound, cell motion is more free or fluid-like. Finally, the “migrated” region includes individual cells that have broken away from the cell monolayer and moved into the middle of the scratch. To analyze vimentin networks in each of these regions, we performed fluorescence imaging (Fig. 5A) and subsequent single-cell analysis, looking at the distribution and polarization of vimentin in individual cells. To aid this distribution analysis, we segmented individual cells into five concentric regions or bins, with bin 1 being perinuclear and bin 5 located near the cell cortex (Fig. 5B). We then examined the fraction of vimentin in each bin (based on mean fluorescence intensity of the vimentin signal) within the jammed, unjammed, and migrated regions (Fig. 5 C and D). We found that PGCCs had a more uniform distribution across the entire cell compared with non-PGCCs, in which vimentin was more tightly clustered around the perinuclear region. Furthermore, we found a consistent contraction of the vimentin area toward the perinuclear region in cells undergoing migration for both cell types. We observed this contractile phenomenon in both the unjammed and migrated regions relative to the jammed region, but this effect was most prominent for cells in the migrated region.

Fig. 5.

PGCCs exhibit differential vimentin organization and increased polarization during directed migration. (A) Representative images of cells located in jammed, unjammed, and migrated regions during the scratch wound assay. MDA-MB-231 cells were stained for DNA (DAPI; cyan), actin (rhodamine phalloidin; red), and vimentin (green). (B) Schematic depicting binning of cell regions, with bin 1 being perinuclear and bin 5 being closest to the cell edge. (C and D) Mean intensity fraction of vimentin at respective bins calculated for non-PGCCs (C) and PGCCs (D). (E) Schematic highlighting vimentin organization during cell migration. (F and G) Vimentin polarization of non-PGCCs and PGCCs in unjammed (F) and migrated (G) regions during directional migration. n > 50 cells for all conditions. All error bars represent SEM. *P < 0.05; **P < 0.01, Student’s t test.

To study vimentin polarization in cell subpopulations, we classified our vimentin distribution into three types (Fig. 5E). The “polarized” vimentin phenotype includes cells with vimentin distributed mainly at the cell’s leading edge (front) or the cell’s trailing edge (behind) relative to the direction of cell movement, which was the direction of the scratch. Cells that had vimentin more evenly distributed throughout the cytoplasm were classified as “unpolarized.” We then examined vimentin polarization of non-PGCCs and PGCCs in the unjammed and migrated regions (Fig. 5 F and G) and found that PGCCs had a much higher incidence of polarized vimentin—more specifically, leading-edge vimentin—compared with non-PGCCs, which had more unpolarized vimentin. This discrepancy was most prominent in the unjammed region but was still significant in migrated regions.

Disruptions to Vimentin Structure Directly Influence PGCC Cell Volume.

To further study the role of vimentin in maintaining the unique PGCC morphology, we disrupted the organization of the VIF network using acrylamide. Previous studies have shown that treating cells with low-dose acrylamide significantly depolymerizes and condenses the vimentin cytoskeletal network around the nucleus or in the perinuclear region (27, 28). Given the role of intermediate filaments in maintaining cell integrity and mechanical stiffness, we hypothesized that treating PGCCs with acrylamide would significantly reduce their volume. To test this hypothesis, we kept both cell types under control conditions or exposed them to 1 mM acrylamide for 12 h (Fig. 6 A and B), then quantified the total intensity (Fig. 6C) and mean intensity (Fig. 6D) of VIFs in PGCCS and non-PGCCs. Acrylamide treatment reduced the intensity of VIFs in both cell types; however, this reduction was significant only for PGCCs, in which the intensity was reduced by ∼30%. Next, we analyzed 3D cell and vimentin volumes using 3D reconstructions based on confocal Z-stacks of control and acrylamide-treated cells (Fig. 6 E and F). Acrylamide treatment significantly decreased the 3D cell volume of PGCCs but not of non-PGCCs (Fig. 6 G and H). The vimentin volume was decreased in both cell types; however, this difference was significant only for PGCCs (Fig. 6 I and J).

Fig. 6.

The network of vimentin intermediate filaments plays a significant role in maintaining PGCC cell volume. (A and B) Representative images of controls (A) and cells treated with 1 mM acrylamide (B). (C and D) Total (C) and mean (D) vimentin intensity of non-PGCCs and PGCCs treated with 0 and 1 mM acrylamide and vimentin siRNA. (E) Representative confocal Z-stack of cells treated with 1 mM acrylamide and stained for DNA (DAPI; cyan), actin (rhodamine phalloidin; red), and vimentin (green). (F) 3D volume reconstruction of the confocal Z-stack of cells treated with 1 mM acrylamide. (G and H) Quantification of cell volume of untreated (n = 36) and 1 mM acrylamide-treated (n = 25) non-PGCCs (G) and untreated (n = 32) and 1 mM acrylamide-treated (n = 29) PGCCs (H) based on confocal Z-stack calculations. (I and J) Quantification of vimentin-occupied volume of untreated (n = 36) and 1 mM acrylamide-treated (n = 25) non-PGCCs (I) and untreated (n = 32) and 1 mM acrylamide-treated (n = 29) PGCCs (J) based on confocal Z-stack calculations. All error bars represent SEM. ns, not significant. *P < 0.05; **P < 0.01, Student’s t test.

Vimentin Disruption Preferentially Inhibits Cell Motility in PGCCs.

Finally, we examined the role of vimentin in cellular unjamming during directed cell migration. In addition to acrylamide treatment, we also used small interfering RNA (siRNA) knockdown of vimentin to ensure that differences in cell motility were due to disruptions in the cell’s VIF network. Cells were transfected with each duplex or with controls using Lipofectamine-3000 (Invitrogen) according to Invitrogen’s recommendations for cell transfection. We repeated our scratch wound healing assay with various concentrations of acrylamide (Fig. 7A and Movie S2), as well as our siRNA knockdown, and found that vimentin disruption was able to inhibit the migration of cancer cells into the scratch wound, with 1 mM acrylamide and siRNA significantly lowering cell migration and 4 mM acrylamide completely blocking migration (Fig. 7B). This effect was very pronounced in polyploidal cells, which showed a significant reduction in the percentage of migrated cells that were PGCCs after both acrylamide and siRNA treatment (Fig. 7C). This suggests that altering or disrupting the vimentin structure of PGCCs could be used to attenuate their migratory persistence.

Fig. 7.

Vimentin disruption blocks directional migration and disrupts vimentin polarization in PGCCs. (A) Scratch wound healing assay of MDA-MB-231 cells treated with 1 mM acrylamide over 12 h. (B) Fraction coverage of cells in scratch wound at 0 and 12 h after scratch, following treatment with 0, 1, and 4 mM acrylamide or vimentin siRNA. (C) Percentage of PGCCs in the migrated region at 12 h after scratch, following treatment with 0, 1, and 4 mM acrylamide or vimentin siRNA. (D) Mean vimentin intensity fraction at various distances from the nuclei for non-PGCCs and PGCCs in the unjammed and migrated regions at 12 h after scratch. (E) Representative image highlighting unpolarized and polarized vimentin organization in migrating cells. (F) Fraction of cells with polarized vimentin treated with 0 and 1 mM acrylamide. (G and H) Normalized change in vimentin polarization of non-PGCCs and PGCCs in unjammed (G) and migrated (H) regions at 12 h after scratch wound assay. n ≥ 500 cells for all motility experiments, and n ≥ 50 cells for all vimentin distribution analysis. All error bars represent SEM. ns, not significant. *P < 0.05; **P < 0.01; ***P < 0.001, Student’s t test.

A distribution analysis of vimentin organization revealed a collapse of the network structure in migrated cells after acrylamide treatment in both cell types. This effect was not observed in cells within the unjammed region, however (Fig. 7D). Interestingly, we observed a consistent reduction in polarized vimentin in PGCCs but not in non-PGCCs (Fig. 7 E and F). Further analysis revealed that targeting vimentin in PGCCs led to a significant reduction in their amount of leading-edge vimentin, resulting in more unpolarized vimentin for PGCCs in both unjammed and migrated regions (Fig. 7 G and H). This reduction in leading edge vimentin likely disrupted PGCC migration, resulting in reduction in their overall motility. This finding highlights the critical role of polarized VIFs in directing the persistent migration of PGCCs.

Discussion

PGCCs are large atypical multinucleated cancer cells often found in late-onset or advanced-stage tumors or after chemotherapy (8, 29). This small population of cancer cells is highly resistant to radiation and chemotherapy and expresses stem cell-like properties important in tumor recurrence after dormancy (8), yet few studies have investigated the role of PGCCs in breast cancer progression to metastasis. In previous work, we identified unique biophysical and transcriptional properties of PGCCs that distinguish them from non-PGCCs (10). PGCCs have increased stiffness, altered cytoskeletal and nuclear structure, and slower but more-directional migration. Our previous study focused largely on the actin cytoskeleton, which plays a critical role in mechanically stabilizing the cell. We showed that targeting actin filaments destabilizes the structure and even sensitizes PGCCs to chemotherapy (10). However, targeting actin, which is ubiquitous to all cells, would be difficult or even impossible in vivo. Furthermore, our previous study did not investigate the underlying mechanism that drove the increased migratory persistence in PGCCs. In the present study, we used single-cell analysis to examine the cytoskeletal structure of PGCCs to elucidate the unique role of VIFs in directing PGCC migration. The reorganization of VIFs serves to facilitate the migration and structural integrity of PGCCs, increasing the migratory persistence and invasive potential of these giant cells.

The abnormal nuclear structure of PGCCs is a commonly described histological feature of human tumors with an abundance of PGCCs often seen in late-stage or drug-resistant tumors (9, 29–33). Recent studies have shown that more aggressive triploid triple-negative breast cancers (TNBCs) have a fourfold higher proportion of PGCCs compared with less aggressive diploid or tetraploid TNBCs (34). Triploid TNBCs (n = 30) were also more resistant to antioxidant therapy with N-acetylcysteine (NAC); PGCCs found in resected tumors from nonresponding patients contained larger numbers of PGCCs that were Ki67- and CD44-positive, indicating that these cells had a more proliferative cancer stem cell-like phenotype (34). Another study showed increased numbers of PGCCs in breast tumors with metastases (n = 52) compared with those without metastases (n = 52) and even fewer PGCCs in benign breast tumors (n = 11) (9); this study also showed that N-cadherin and vimentin were both dramatically up-regulated in more advanced tumors and in PGCCs with budding daughter cells (9). An abundance of PGCCs also has been reported in high-grade and metastatic ovarian cancers compared with primary ovarian tumors without metastases and borderline serous ovarian tumors; these PGCCs were also found in hypoxic tumor regions and expressed stem cell-like markers important in tumor relapse and chemoresistance (29).

In this study, we focused on vimentin, which plays a fundamental role in regulating cell architecture and function (35), determining nuclear positioning and shape (36), and regulating chromatin condensation and gene expression (36–38). Our single-cell analysis of vimentin-stained PGCCs helped elucidate the enhanced role of VIFs in supporting their structural integrity and enlarged morphology. We found that PGCCs express higher vimentin gene expression levels, as corroborated by immunofluorescence staining. PGCCs are under inherent stress from their enlarged volume and require certain adaptations to maintain the mechanical integrity of their cytoskeleton. Numerous studies have highlighted the unique role of VIFs in regulating cell shape and structure (20, 28), as well as in responding to shear stress (39, 40). Our 3D analysis of vimentin distribution reveals a more diffuse network of VIFs throughout the cytoplasm in PGCCs as opposed to the perinuclear bundling observed in non-PGCCs. This suggests greater involvement of VIFs in regulating the structural support of PGCCs. When we disrupted the VIF network using acrylamide, we saw significant reductions in overall cell volume. This corroborates previous reports from the Janmey laboratory of a significant reduction in the cell volume of vimentin knockout mouse embryonic fibroblasts compared with controls and rescues (18).

Vimentin can protect the nucleus and organelles from repeated strain (20). For example, when migrating cells undergo extreme nuclear deformation in confined spaces, they rely on VIFs to buffer the nucleus and protect against DNA damage and apoptosis (18). As cells proliferate in a constrained physical dimension akin to proliferation of cells within a tumor, cells begin to experience “caging” from surrounding cells. Once the crowding reaches a certain threshold, the cells are jammed in place, and the collective cell sheet behaves like a solid (26). To become unjammed and migrate more freely, a cell must squeeze through constrained spaces and past its neighbors, which often necessitates large nuclear and cytoplasmic deformations. Given that the ability of cells to become motile depends directly on their physical interactions with other cells, it is logical to assume that cell rheology is a key determinant of whether an individual cell becomes unjammed. Owing to the unique hyperplastic properties of the VIF network (19, 20), vimentin plays a predominant role in the viscoelastic deformation of migrating cells and fulfills a protective role as these cells squeeze past their neighbors (18, 20). It is likely that the cytoprotective role of vimentin is even more important in 3D environments, where cells undergo large deformations to squeeze through tight spaces in the surrounding matrix.

The protective nature of VIFs is especially important in PGCCs, which undergo more strain during migration due to their greater nuclear and cytoplasmic volume. Our data show that polarized VIFs provide structural support for PGCCs while directing their migration from the jammed region to the migrated region of the scratch wound. This persistent migration is associated with highly polarized organization of VIFs in the direction of cell motion. When these VIF networks were disrupted by siRNA knockdown of vimentin or by collapsing the structure using acrylamide, PGCCS exhibited a dramatic reduction in cell motility. The reduced migration into the scratch wound was due to the loss of polarization in vimentin and subsequent lack of polarity in cell movement. This finding suggests that compared with non-PGCCs, PGCCs rely more heavily on their VIF network to maintain their polarity. One explanation for this is that because intermediate filaments are relatively stable, PGCCs use vimentin to resist axial strain and to align their multinucleated structure along the axis of migration. Furthermore, VIFs are directly coupled to the actin and microtubule networks that mechanically support the unique biophysical properties of PGCCs (10, 17). Vimentin could play an important role in aligning these cytoskeletal networks through cross-linking proteins, such as plectin, to aid cell polarity and migratory persistence.

Overall, the polarization of vimentin results in polarization of cell migration, increasing their migratory potential owing to higher persistence in movement (41, 42). This association of vimentin polarization to cell migratory behavior corroborates previous studies demonstrating that polarized vimentin serves as a template for microtubule plus end growth and also promotes actin treadmilling, facilitating directionality in cell migration (16). The polarization of vimentin can be disrupted using destabilizing agents such as acrylamide and siRNA knockdown, resulting in a significant reduction of PGCC motility. This effect likely also can be achieved through knockdown or disruption of vimentin/actin cross-linking proteins, such as plectin, which present a much more effective therapeutic target for application in the clinic.

Given our findings regarding the increased dependence of polyploidal cells on their vimentin network, targeting vimentin or vimentin cross-linking proteins could provide a novel therapeutic strategy for targeting PGCCs in the clinic. It has long been established clinically that the expression of vimentin in tumors is linked to more aggressive and metastatic cancers, although the exact mechanism is unknown. PGCCs rely heavily on their VIF network to maintain their enlarged morphology and to direct their persistent migration (38, 43). In addition, this VIF network affects nuclear arrangement, which is critical in amitotic budding to give rise to daughter cells (44). By inhibiting PGCC budding, we could minimize the impact of PGCCs in cancer progression, as the formation of daughter cells is a key aspect of PGCC tumorigenesis. PGCCs are increasingly implicated in tumor relapse after therapy and chemoresistance (5, 6, 11), which leads to poor patient outcomes (43). By furthering our understanding of the underlying cytoskeletal features driving the migratory persistence of PGCCs, we hope to shed light on their unique role in cancer progression to metastasis.

Materials and Methods

Cell Culture.

MDA-MB-231 human breast cancer cells and MDA-MB-453 (American Type Culture Collection) were cultured in DMEM/Ham’s F-12 50/50 mix (Corning). SKBR-3 was cultured in McCoy’s 5A (modified) medium. The cell culture medium was further supplemented with 10% FBS (Atlanta Biologicals) as well as 1% penicillin-streptomycin (Corning). Cells were maintained at 37 °C in 5% CO₂ until the time of experiment/fixation.

qRT-PCR Analysis.

mRNA for genetic analysis was extracted from MDA-MB-231 PGCCs and non-PGCCs using PureZOL RNA isolation reagent (Bio-Rad) according to the manufacturer’s recommendations. RNA integrity and purity were confirmed by spectrophotometry (Nanodrop 1000; Thermo Fisher Scientific). gDNA removal and cDNA synthesis were performed using the iScript gDNA Clear cDNA Synthesis Kit (Bio-Rad) according to the manufacturer’s protocol. Here 1 μg of RNA was converted into cDNA for real-time PCR analysis of EMT markers. The primers used are listed in SI Appendix, Table S1.

Scratch Wound Assay.

MDA-MB-231 cells were seeded at high concentration and allowed to grow to 100% confluency in a 24-well plate (Corning) (SI Appendix, Fig. S2). The well plates were pretreated with 20 μg/mL type I rat tail collagen (Corning) for 1 h before seeding. Before imaging, confluent monolayers were scratched vertically and horizontally in a cross fashion with a 10-μL pipette tip. The well was then washed three times with PBS and fed using medium with Hoechst 33342 (1 μg/mL) for an additional 30 min. Hoechst-stained cells were then imaged with a Nikon Eclipse Ti inverted fluorescent microscope equipped with an environmental chamber (37 °C and 5% CO₂). Edges of scratch wounds were selected and imaged periodically every 10 min for 12 h. Time-lapse videos were then exported, and individual images were fed into CellProfiler (Broad Institute). Using a custom pipeline, individual cells were identified within migrated regions. In brief, we used object segmentation after thresholding at 0 h to identify individual nuclei. Identified nuclei were then dilated to estimate the region of the initial monolayer (i.e., the initial coverage area). We repeated this process at different intervals to establish a time course, allowing us to determine the coverage ratio as well as to isolate migrated populations. PGCCs were subsequently identified by measuring the area of segmented nuclei.

Immunofluorescence Staining.

Cells designated for imaging at high magnification or by confocal microscopy were seeded in 24-well plates containing 12-mm, 1.5-μm-thick glass coverslips (Electron Microscopy Sciences). Scratch wounding in a cross formation was performed using a 10-μL pipette tip, and cells were allowed to migrate before fixation for predetermined times. Cells were fixed using 4% formaldehyde and 1% Triton X-100, then stained for DNA, actin, and vimentin cytoskeletal networks as described previously (1–3). In brief, a 1:100 dilution of anti-rabbit vimentin antibody (Cell Signaling Technology) was added to individual coverslips for 1 h. After incubation with primary antibody, coverslips were returned to well plates and then washed twice with PBS. Next, a 1:1,000 anti-rabbit Alexa Fluor 488 secondary antibody (Invitrogen) was used in conjunction with rhodamine-phalloidin (0.2 uM) and DAPI (5 ug/mL) for 1 h. Coverslips were washed three times with PBS and then mounted on microscope slides using Fluoromount-G (SouthernBiotech).

Images of the scratch wound were divided into migrated, unjammed, and jammed regions, defined as follows. The migrated region includes cells that have migrated and moved into the initial scratch area. The unjammed region represents cells located near the edge of the scratch wound, which are uncaged/unjammed by their neighbors and can migrate freely. Finally, the jammed region includes areas toward the middle of the monolayer, where cells are caged/jammed by their neighbors and cannot move freely. Images of fluorescently labeled cells were analyzed for vimentin polarity. In brief, the direction of cell migration was determined primarily by the position of the cell with respect to the scratch wound and overall cell morphology (i.e., identification of axis of elongation and lamellipodia structures). Vimentin polarity was then assessed with respect to the direction of cell migration (SI Appendix, Fig. S3).

Fluorescence Imaging and Volume Analysis.

Mounted slides were imaged using a Nikon Eclipse Ti inverted fluorescent microscope with a 60× objective or an Olympus 3000 confocal microscope (Z-stacks for 3D volume reconstructions). Images from confocal microscopy were analyzed with ImageJ using the Voxel Counter plugin after thresholding and gap closing to determine cell volume. In brief, confocal Z-stacks of phalloidin-stained actin were thresholded by slice, then a gap closing function was used to fill holes within the thresholded area (to represent cell area per slice). Finally, the Voxel Counter plugin was used to calculate the number of voxels included in the thresholded area, to measure overall cell volume. A similar approach was applied on fluorescently stained vimentin and Hoechst-stained nuclei, and relative vimentin/nuclei volume was determined in relation to the total volume of the cell. 3D projections were generated using the 3D viewer functionality in Fiji.

siRNA Treatment.

MDA-MB-231 vimentin was silenced using a pool of three siRNA duplexes obtained using the TriFECTa DsiRNA Kit for vimentin (Integrated DNA Technologies). Cells were transfected with each duplex or with controls using Lipofectamine-3000 (Invitrogen) according to the manufacturer’s recommendations. Knockdown efficiency was validated using qRT-PCR expression analysis of cells transfected with the silencing duplex compared with the provided scramble control. The absence of off-target effects was confirmed using the siRNA duplex and assessing the expression level of HPRT1 (control gene not targeted by the duplexes) compared with HPRT1 knockdown duplex control.

Data Availability.

All study data are included in the main text and SI Appendix.