α-Actinin-4 drives invasiveness by regulating myosin IIB expression and myosin IIA localization

Cancer metastasis, that is, spread of cancer cells from one tissue to another, is the primary reason behind the high mortality of cancer. Consequently, identifying molecules mediating cancer metastasis and their mechanisms of action can provide us with useful strategies for therapeutic intervention. Most epithelial cancers are marked by dramatic alterations in composition and organization of the extracellular matrix (ECM) (Lu et al., 2012; Pickup et al., 2014). ECM stiffening, induced by excessive deposition of collagen I and its crosslinking by lysyl oxidase reported in breast cancer, has been shown to be capable of driving cancer progression (Provenzano et al., 2008; Wang et al., 2017). For successful invasion through these dense matrices, cancer cells are known to upregulate expression of matrix-degrading enzymes such as matrix metalloproteinases (MMPs) for locally degrading the matrix and creating paths for escape (Itoh and Nagase, 2002; Kessenbrock et al., 2010; MacDougall and Matrisian, 1995). However, even after such localized degradation, translocation of the nucleus, which is large and stiff, represents a rate-limiting factor for confined migration (Davidson et al., 2014). Whereas non-muscle myosin IIA (NMM IIA; heavy chain encoded by MYH9) is essential for active force generation during protrusion, nuclear translocation is largely mediated by NMM IIB (heavy chain encoded by MYH10), which exhibits perinuclear localization and mediates nuclear squeezing (Thomas et al., 2015).

α-Actinin-4 (ACTN4) is a non-muscle actin crosslinking protein whose expression has been strongly correlated with cancer invasiveness, metastasis and therapeutic resistance across different types of cancers (Desai et al., 2018; Sen et al., 2009; Honda et al., 1998; Honda, 2015; Hsu and Kao, 2013; Wang et al., 2015). These behaviors are attributed to involvement of ACTN4 with several well-known signaling molecules, including β-catenin (An et al., 2016), NFκB (Aksenova et al., 2013) and AKT proteins (Desai et al., 2018). Recently, ACTN4 along with filamin A, NMM IIA and IIB were identified as four mechanoresponsive proteins associated with the actin cytoskeleton that were enriched at sites of localized pressure (Schiffhauer et al., 2016). Although these studies illustrate the importance of ACTN4 as a central player regulating several aspects of cancer invasion, the mechanisms by which it contributes to increased invasiveness remains incompletely understood.

In this study, we have specifically probed the role of ACTN4 in regulating cancer invasiveness. Using the TCGA (Cerami et al., 2012; Gao et al., 2013; Weinstein et al., 2013) database and breast cancer cell lines of varying invasiveness, we first establish a positive correlation between ACTN4 levels and cell motility. We then use MDA-MB-231 breast cancer cells and HT-1080 fibrosarcoma cells, which express a high level of ACTN4, to probe the functions of ACTN4 in regulating cell migration and invasion. We show that ACTN4 knockdown leads to cell elongation and formation of smaller focal adhesions, but a reduction in protrusion–retraction dynamics. Using collagen gel invasion and transwell pore migration experiments, we then show that ACTN4 regulates nuclear translocation by modulating NMM IIB expression. Our experiments further establish an indirect association of ACTN4 with NMM IIA mediated by a functional F-actin network at the lamellipodia–lamellum interface. This association is essential for peripheral localization of NMM IIA. Finally, using a computational model, we estimate ACTN4-mediated protrusive forces at the cell front required for confined migration. Collectively, our results suggest that ACTN4 regulates cancer invasiveness by regulating nuclear deformation via modulation of NMM IIB expression and by stabilizing protrusions at the cell front via its association with NMM IIA.

As per TCGA database studies (Cerami et al., 2012; Gao et al., 2013; Weinstein et al., 2013), higher ACTN4 expression in different types of cancers is associated with lower overall survival as well as lower progression-free survival (Fig. 1A). Within different breast cancer sub-types, higher ACTN4 expression was associated with lower disease-free survival in triple negative breast cancers (TNBCs), although the results were not statistically significant (Fig. S1A). In contrast, in estrogen receptor (ER)-positive, progesterone receptor (PR)-positive and HER2-negative breast cancer patients, as well as in HER2-positive breast cancer patients, higher ACTN4 led to better disease-free survival. In comparison to ACTN4, no correlation was observed between α-actinin-1 (ACTN1) expression and survival (Fig. S1B). Since cancer mortality is linked to cancer metastasis, to probe the role of ACTN4 in regulating invasiveness of breast cancer cells, ACTN4 levels and localization were probed in MCF-7, T47D, ZR-75 and MDA-MB-231 breast cancer cell lines. Of these, MCF-7 cells are tumorigenic but non-invasive, and TNBC MDA-MB-231 (hereafter MDA) cells are highly metastatic. In addition, invasive HT-1080 (hereafter HT) fibrosarcoma cells were also considered. MCF-7, T47D and ZR-75 cells, which possessed low ACTN4 levels (Fig. 1B), were also the least motile (Fig. 1C,D; Movie 1), with ACTN4 exhibiting a primarily cytoplasmic localization (Fig. 1E). In comparison, MDA and HT cells, which possessed highest levels of ACTN4, were 5–6-fold more motile than the above cell types. In these invasive cells, ACTN4 was prominently localized at the cell periphery as evidenced by the clear enrichment at the cell periphery (Fig. 1E). Higher ACTN4 expression levels also correlated with higher expression of β1 integrins and the mesenchymal marker vimentin (Fig. 1B). Collectively, these results establish a clear correlation between ACTN4 levels and its membrane localization with cancer motility.

To probe the mechanisms of regulation of cancer invasiveness by ACTN4, two stable ACTN4 knockdown lines (shAC#1 and shAC#2) were established for both MDA and HT cells, with ACTN1 levels unchanged in ACTN4-knockdown cells (Fig. 2A–C). Knockdown cells were found to be more elongated (i.e. less circularity) compared to control cells (shCTL), with an increase in spreading observed in MDA cells, but not in HT cells (Fig. S2A,B). However, these changes were not associated with alterations in the expression of epithelial-to-mesenchymal transition (EMT) markers integrin β1 and vimentin (Fig. 2C; Fig. S2C). ACTN4 knockdown led to marginal reduction in proliferation rate of MDA cells, but not of HT cells (Fig. S2D).

In line with its crosslinking function, ACTN4 knockdown led to a marked reduction in the number of stress fibers (Fig. S2E,F). Increase in the average filament length in knockdown cells can be attributed to the disappearance of shorter actin filaments from the cell leading to increase in average filament length (Fig. S2G). Consistent with the reduction in the number of F-actin filaments, cortical stiffness probed with AFM revealed that knockdown cells were softer than controls (Fig. S2H,I).

Loss of ACTN4 led to reduced motility in both MDA and HT cells, with a more pronounced effect in MDA cells (Fig. 2D,E; Movies 2,3). Kymograph analysis of leading-edge protrusion dynamics revealed the presence of protrusion–retraction (P-R) cycles similar to that observed in other adherent cell types (Fig. 2F). Both the frequency of P-R cycles as well as the protrusion and retraction distances covered in each P-R cycle were reduced in knockdown cells, leading to reduction in both protrusion rate and retraction rate (Fig. 2G).

The transition from the end of one cycle of retraction to initiation of another protrusion cycle coincides with formation of focal adhesions at the lamellipodia–lamellar interface (Ponti et al., 2004). Quantification of paxillin-positive focal adhesions revealed a prominent reduction in both the number and size of focal adhesions in ACTN4 knockdown cells (Fig. 2H,I). Further characterization of paxillin turnover at focal adhesions revealed that adhesion lifetime was increased in knockdown cells (Fig. 2J,K; Movies 4,5). Together, these results establish ACTN4 as a key molecule regulating cytoskeletal organization and adhesion dynamics, both of which are relevant to invasion.

To next probe the importance of ACTN4 in regulating cancer invasiveness through 3D matrices, cells were embedded in 3D collagen gels with mean pore sizes of ∼3 µm for 1.2 mg/ml gels (used for MDA cells) and ∼2 µm for 1.5 mg/ml gels (used for HT cells) (Fig. 3A). While long ACTN4-enriched protrusions were observed in control cells (Fig. S3A), a drop in both the number of protrusions and the average protrusion length in knockdown cells (Fig. 3B,C) led to a marked reduction in the cell speed of ACTN4 knockdown cells (Fig. 3D,E; Movies 6,7).

Nuclear deformation represents a rate-limiting factor for migration in 3D matrices, particularly through sub-nuclear sized pores (Davidson et al., 2014). To probe the role of ACTN4 in mediating nuclear deformation, transwell migration experiments were performed using 3-µm size transwell pores (Fig. 3F). Comparison of the average number of nuclei at the top surface and bottom surface of the pores allowed us to compare the translocation efficiency of control and knockdown cells. In line with the above 3D motility findings, comparison of the fraction of cells that successfully transited through the transwell pores (i.e. the translocation efficiency), revealed a dramatic drop in knockdown cells (Fig. 3G,H).

A lower translocation efficiency of ACTN4 knockdown cells was associated with an increase in the average nuclear height of knockdown cells (Fig. 3I). Increased nuclear height of knockdown cells was associated with reduced contractility of knockdown cells, as assessed with a gel compaction assay (Fig. 3J). Consistent with this, treatment of control cells with the myosin inhibitor blebbistatin (Blebb) led to drop in motility to levels comparable to that of knockdown cells (Fig. 3K; Fig. S3B). Although protrusions were still observed in Blebb-treated cells, inhibition of actin polymerization by cytochalasin D (CytoD) led to complete cell rounding and near complete loss of motility. Together, these observations suggest that ACTN4 regulates invasiveness by modulating protrusion dynamics and actomyosin contractility.

Efficiency of nuclear translocation through pores is dictated by physical properties of the nucleus, including size and stiffness, as well as by the levels of non-muscle myosin IIB (NMM IIB), which plays a key role in mediating nuclear deformation (Thomas et al., 2015). In line with the literature, NMM IIB was detected spanning the top of the nucleus in control cells (Fig. 4A), and was found to colocalize with the peri-nuclear actin network visualized by staining cells with ACTN4 and F-actin (Fig. S4A). Since the average nuclear height of knockdown cells was found to be greater (Fig. 3I) without any alteration in lamin A/C levels and its phosphorylation (Fig. S4B,C), we checked whether myosin levels and/or activity are perturbed in knockdown cells. Expression profiling of NMM IIA and IIB revealed a dramatic reduction in NMM IIB (MHY10) levels both at transcriptional level and at protein level in knockdown cells (Fig. 4B,C; Fig. S4D). Increase in 3D motility of non-malignant MCF7 cells overexpressing ACTN4 further establishes the importance of ACTN4 in cell motility (Fig. S4E,F, Movie 8). However, ACTN4 overexpression did not lead to NMM IIB upregulation suggesting that ACTN4-mediated regulation of NMM IIB expression might be cell type specific (Fig. S4G).

In comparison to the reduction in NMM IIB expression in ACTN4 knockdown cells, NMM IIA (MHY9) expression remained unperturbed in knockdown cells. However, there was a pronounced reduction in the activation of myosin light chain phosphorylation (pMLC; herein referring to phospho-myosin light chain 2) at the cell periphery (Fig. 4E,F). In contrast to the peripheral localization of NMM IIA in control cells, NMM IIA exhibited primarily cytoplasmic localization in knockdown cells (Fig. 4D). Similar localization dynamics was observed in control and knockdown cells transiently transfected with eGFP–NMM IIA (Movies 9,10). To test whether cytoplasmic localization of NMM IIA in ACTN4 knockdown cells is a compensatory response to NMM IIB downregulation, we overexpressed NMM IIB in knockdown cells (Fig. S5A). However, no change in localization of NMM IIA was observed in these cells, suggesting that peripheral NMM IIA retention is independent of NMM IIB expression (Fig. S5B).

To next test whether ACTN4–NMM IIA association or interaction is necessary for NMM IIA retention at the cell periphery, co-immunoprecipitation (co-IP) and proximity ligation assays (PLA) were performed. These experiments revealed a close association between NMM IIA and ACTN4 in control cells (Fig. 4G; Fig. S5C), with the cell periphery as the preferred location of association of these two molecules (Fig. 4H; Fig. S5D). Co-staining of cells for NMM IIA and the lamellipodial marker cofilin revealed the coexistence of regions where NMM IIA intensity peak trailed the cofilin intensity peak as well as regions where the peaks colocalized at the cell periphery (Fig. S5G,H). Live imaging of cells co-transfected with eGFP–NMM IIA and mCherry–cofilin1 revealed the presence of cofilin and absence of NMM IIA in the lamellipodial regions (Fig. S5I,J, Movie 11). These observations confirm the absence of NMM IIA in the lamellipodia (Ponti et al., 2004), and suggest the lamellipodia–lamellum interface as the probable site of ACTN4–NMM IIA colocalization.

Interestingly, the extent of ACTN4–NMM II A association dropped by ∼50% upon cytochalasin D (CytoD)-induced F-actin depolymerization. Immunostaining revealed a similar drop in the extent of colocalization of ACTN4–NMM IIA in CytoD-treated cells (Fig. S5E,F). The extent of ACTN4–NMM IIA colocalization was comparable to that of NMM IIA–F-actin colocalization levels in CytoD-treated cells and in vehicle (DMSO)-treated ACTN4 knockdown cells. Furthermore, CytoD treatment did not alter the extent of NMM IIA–F-actin colocalization in knockdown cells. These results suggests that ACTN4–NMM IIA association may be indirect and is mediated by an intact F-actin network.

To check whether ACTN4 binding to F-actin alters NMM IIA binding, preliminary docking studies were performed using the ClusPro protein docking server (Kozakov et al., 2017), wherein NMM IIA was docked to either a 6-monomer F-actin filament alone or to ACTN4-bound F-actin (Fig. S6A). NMM IIA docking onto ACTN4-bound F-actin was performed for two different configurations of NMM IIA and ACTN4 relative to F-actin. While NMM IIA and ACTN4 were docked on adjacent actin monomers in the first case (Dock #1), in the second case, the distance between NMM IIA and ACTN4 spanned a 3-monomer distance (Dock #2). Although NMM IIA–F-actin binding energy was not significantly altered when NMM IIA and ACTN4 were bound to F-actin, the largest cluster size (indicative of more stable binding) was observed for the case when NMM IIA and ACTN4 were spaced a few monomers apart (Fig. S6B). Collectively, our results suggest that the invasive phenotype of cancer cells is maintained by ACTN4 through regulation of NMM IIB expression and an indirect association with NMM IIA at the lamellipodia–lamellum interface.

To finally establish that the observed changes in knockdown cells are a direct consequence of ACTN4 knockdown, knockdown cells were transiently transfected with eGFP–ACTN4 yielding a transfection efficiency of ∼20%. When embedded in 3D collagen gels, eGFP–ACTN4-transfected cells were found to migrate faster than knockdown cells and also exhibited increased protrusion formation (Fig. 5A,B; Movie 12). Comparison of the proportion of transfected cells between the top and bottom of the pores revealed a 4–5-fold enrichment of eGFP–ACTN4-transfected cells at the bottom of the pores (Fig. 5C,D) with eGFP–ACTN4 detected at sites of pore entry (Fig. 5E). In comparison, in pcDNA-eGFP-transfected control cells, no enrichment was observed at the bottom of the pores (Fig. S7A,B). Furthermore, in knockdown cells transfected with eGFP–ACTN4, peripheral localization of NMM IIA and its association with ACTN4 were restored (Fig. 5F). Activated myosin (i.e. pMLC) was also found to colocalize prominently with ACTN4 at the cell periphery (Fig. 5G). Taken together, our results establish ACTN4 as a key regulator of cytoskeletal organization and nuclear deformability through modulation of NMM IIA localization and NMM IIB expression.

Our results suggest that ACTN4 regulates invasion by regulating protrusion dynamics at the leading edge, and mediating nuclear deformation by regulating NMM IIB expression. To estimate ACTN4-dependent forces required for invasion, we developed a finite element (FE) based computational model of confined cell migration accounting for physical properties of the cell and the nucleus (refer to the Materials and Methods for details). In our model, cell entry into a pore was mediated by protrusive forces F_P at the cell front, with higher F_P associated with higher ACTN4 levels and increased protrusive force generation (Fig. 6A). The region of generation of F_P is representative of localization of NMM IIA at the cell front (in the direction of migration, depicted in blue in Fig. 6A) and its association with ACTN4. The NMM IIB-mediated squeezing of the nucleus is encoded in the model as a consequence of FE mesh spring compression under external matrix-induced stresses during confined migration.

Simulations were performed for varying combinations of matrix stiffness (E_m=1–5 kPa) and pore sizes (ϕ=3–8 μm), while keeping nucleus stiffness (E_n) constant at 5 kPa (Das et al., 2019). It was found that a decrease in pore size led to an increase in cell elongation and time to migrate through the pore for E_n=E_m=5 kPa (Fig. 6B). Cell elongation, measured as a ratio between the cortico-nuclear distance (X_CN) at time instant ‘t’ to that at t=0 [i.e. X_CN(t)/X_CN(0)] was found to increase with decrease in F_Pmax and E_cortex, in concordance with our experimental observations (Fig. 6C). Correspondingly, the force F_p predicted by our model to be generated by a cell to enter a pore increased with increase in matrix stiffness and decrease in pore size (Fig. 6D). Since low ACTN4 density, and NMM IIA and IIB expression correspond to low predicted cell-generated forces, the time required by cells to migrate through confined pores subsequently increases (Fig. 6E). In simulations with very low F_Pmax, denoting negligible ACTN4 levels, cells do not even migrate within the maximum computational time duration of 6000s, further confirming our experimental observation of stalling of cell migration.

It is interesting to note from our simulations that although the predicted F_P is considerably higher for 3 μm pores than 5 μm pores, the migration time required is similar for cells with low ACTN4 levels migrating through 5 μm pores (blue dotted line) and those with moderate ACTN4 levels migrating through 3 μm pores (red dashed line) (Fig. 6E). Average cell speeds increase with pore size, but decrease with a decrease in ACTN4 levels and cortical stiffness (Fig. 6F). Average cell speeds in the range 6–54 μm/h, as predicted by our model, are comparable to experimental observations. Average cell speeds through large pores (ϕ=8 μm) remain relatively unchanged whereas, small or moderate pores led to an increase in cell speed with E_n/E_m.

The change in nuclear circularity predicted by our model during cell entry into a pore decreases maximally for ϕ=3 μm and E_n/E_m=1 (Fig. 6G). A small increase in nuclear circularity is also observed for ϕ=5 μm and E_n/E_m=5 for low versus normal ACTN4 levels which might be attributed to decrease in NMM IIB-mediated contractility in conjunction with ACTN4 depletion. Although our computational model makes some assumptions, these model results match well to our experimental observations. Taken together, they suggest that ACTN4 modulates cell migration by regulating the expression and localization of NMM IIA and IIB.

Our findings implicate ACTN4 as a regulator of cancer invasiveness via its interaction with NMM IIA at the lamellipodia–lamellum interface and through regulation of NMM IIB expression (Fig. 7). Consistent with previous studies (Feng et al., 2013; Meacci et al., 2016; Oakes et al., 2012), in both MDA-MB-231 and HT-1080 cells, ACTN4 knockdown led to reduction in the number of F-actin stress fibers and cortical softening. In contrast, in MDCK cells, ACTN4 knockdown has been associated with increased F-actin content. Increased stress fiber formation in these cells has been linked with increased F-actin association with tropomyosin, and is indicative of a competition between ACTN4 and tropomyosin for binding to F-actin (Kemp and Brieher, 2018). The lack of F-actin stabilization in MDA-MB-231 cells may be attributed to the absence or low expression of tropomyosin isoforms in these cells (Varga et al., 2005).

In addition to its actin crosslinking function, ACTN4 is increasingly known to regulate several aspects of cancer including cell proliferation (Huang et al., 2019; Khurana et al., 2011), EMT (An et al., 2016) and therapeutic resistance (Desai et al., 2018; Jung et al., 2019). EMT is associated with transition from an epithelial cell–cell adhesion-rich state to a mesenchymal state characterized by formation of cell–matrix adhesions. In both MDA-MB-231 and HT-1080 cells, which are mesenchymal in nature, the loss of paxillin-positive focal adhesions and increase in adhesion lifetime in ACTN4 knockdown suggests that ACTN4 is important for both adhesion formation and adhesion turnover. The inability of ACTN4 to bind zyxin in colorectal cancer cells has been implicated in preventing adhesion maturation, thereby enabling faster turnover (Fukumoto et al., 2015). In melanoma cells that express high levels of ACTN4 and exhibit amoeboidal migration, ACTN4 knockdown leads to transition from an amoeboidal morphology to a more mesenchymal state via an increase in focal adhesion size (Shao et al., 2014). These results thus suggest that ACTN4 levels dictate the mode of invasion (i.e. mesenchymal versus amoeboidal) by modulating the extent of cell–matrix adhesion.

Migrating cells consist of lamellipodia and lamella with the former composed of a branched actin network, and the latter enriched in more bundled actin (Koestler et al., 2008; Vallotton and Small, 2009). While nascent focal adhesions are formed in the lamellipodial region in a force-independent manner, their maturation is brought about by coupling the actomyosin cytoskeleton to focal adhesions. Both α-actinin and NMM IIA have been shown to play prominent roles in focal adhesion maturation, with actin crosslinking function of α-actinin responsible for locally organizing the actin network, and NMM IIA-mediated contraction of the actin network mediating focal adhesion stabilization (Choi et al., 2008). ACTN4 has been reported to participate in adhesion maturation by formation of stress fibers (Hotulainen and Lappalainen, 2006; Triplett and Pavalko, 2006). Parallelly, NMM IIA is also required for growth of focal adhesions, with adhesion maturation impaired in NMM IIA deficient cells (Doyle et al., 2012; Vicente-Manzanares et al., 2007). Our studies suggest that ACTN4-mediated organization of the F-actin network is essential for retention of NMM IIA at the cell periphery and its phosphorylation. Since NMM IIA is not present in the lamellipodial regions, based on NMM IIA–cofilin localization patterns observed both in fixed cells and in live-cell imaging, we posit that ACTN4–NMM IIA association occurs at the lamellipodia–lamellum interface where focal adhesions mature (Ponti et al., 2004). Loss of peripheral NMM IIA and reduction in focal adhesion size highlights the importance of ACTN4 in mediating retention of NMM IIA at anterior lamellar regions and maturation of focal adhesions by traction forces. The increase in adhesion lifetime in ACTN4 knockdown cells may be associated with NMM IIA association with ACTN1, which is also known to link F-actin to focal adhesions (Rajfur et al., 2002), exhibits contrasting behavior compared to ACTN4 (Quick and Skalli, 2010) and has significantly higher actin binding affinity in comparison to ACTN4 (Ferrer et al., 2008). In cells expressing high levels of ACTN1, similar to ACTN4–NMM IIA interactions, ACTN1–NMM IIA interactions may regulate protrusion dynamics by modulating focal adhesion turnover. However, given differences between ACTN1 and ACTN4 in regulating focal adhesion dynamics (Fukumoto et al., 2015), it is likely that ACTN1–NMM IIA association will lead to more stable adhesions and slower turnover.

What is the mechanism by which ACTN4 retains NMM IIA at the cell front? Although co-IP studies revealed an association between these two proteins, the drop in their association in the presence of the actin-depolymerizing drug CytoD suggests that the interaction is indirect and mediated by a functional F-actin cytoskeleton. Comparable levels of colocalization of NMM IIA with ACTN4 and F-actin in both control and CytoD-treated cells are also indicative of an indirect association between NMM IIA and ACTN4. Molecular docking studies suggest that NMM IIA and ACTN4 binding on F-actin a few monomers apart is more favorable than NMM IIA binding to F-actin alone. These results suggest that retention of NMM IIA at the lamellipodia–lamellum interface can be attributed primarily to ACTN4-mediated F-actin bundling and its organization. However, the role of ACTN4 in spatial regulation of NMM IIA phosphorylation by ROCK and MLCK pathways remains to be established.

When embedded in collagen matrices, MDA-MB and HT cells exhibit elongated morphologies with prominent localization of ACTN4 and F-actin along the protrusions. Loss of protrusions in ACTN4 knockdown cells suggests that leading edge protrusion is driven by polymerization of F-actin filaments stabilized by ACTN4 crosslinks. Under these conditions, physical resistance provided by the surrounding tissues may not only drive increased accumulation of ACTN4 given its mechanoresponsive nature, but also lead to increased actin binding, thereby stabilizing the protrusions (Schiffhauer et al., 2016; Thomas and Robinson, 2017). Our computational model predicts that during migration through larger pores (ϕ=8 μm), protrusive forces (F_P) required for pore entry and cell speeds are nearly constant over a wide range of E_n/E_m values (1−5), which is consistent with the findings of statistical mechanics-based confined migration model (Kumar et al., 2018). While migrating through increasingly smaller pores, even for sustaining moderate cell motility (10−20 μm/h), 10–20-fold higher protrusion forces (F_P) are required, highlighting the importance of the mechanoaccumulative nature of ACTN4 required for mediating confined migration (Schiffhauer et al., 2016). Reduction in gel compaction and peripheral myosin phosphorylation in ACTN4 knockdown cells, and the drop in motility of Blebb-treated control cells to levels comparable to that of ACTN4 knockdown cells highlights the role of ACTN4 in driving confined migration via regulation of protrusion dynamics and actomyosin contractility. Our results predict that protrusive forces, F_P, scale with the degree of cell confinement (i.e. decreasing pore size ϕ) and with increasing ECM stiffness (i.e. decreasing E_n/E_m). This predictive relationship can be utilized to estimate the protrusive force required by a cell to migrate through a pore in a matrix of given stiffness.

The nucleus, which is large and stiff, represents the rate limiting factor for 3D migration (Davidson et al., 2014; Harada et al., 2014; Rowat et al., 2013; Wolf et al., 2013). NMM IIB, which exhibits peri-nuclear localization, has been shown to be critical for mediating nuclear deformation and translocation by physical coupling the actomyosin cytoskeleton to the nucleus via the LINC complex (Halder et al., 2019; Kapoor et al., 2018; McGregor et al., 2016; Thomas et al., 2016). Our observations of increase in nuclear height and concomitant loss of NMM IIB in ACTN4 knockdown cells are consistent with the above described function of NMM IIB (Meshel et al., 2005; Thomas et al., 2016). Reduction in NMM IIB both at protein level and mRNA level are indicative of ACTN4 being a transcriptional regulator of NMM IIB. Thus, upregulation of NMM IIB in cells undergoing EMT (Beach et al., 2011) might be driven by ACTN4, which is an EMT inducer. Although ACTN4 has been reported to translocate to the nucleus (Kumeta et al., 2010), the mechanism(s) by which ACTN4 regulates NMM IIB expression remain(s) to be established. ACTN4 induces EMT by activating AKT proteins and inducing expression of EMT associated transcription factors such as Snail and Slug (An et al., 2016). Also, ACTN4 has been identified as a co-activator of several transcription factors including p65 subunit of NFκB (Aksenova et al., 2013). However, the precise mechanisms of how ACTN4 regulates expression of multiple genes relevant to EMT, including that encoding NMM IIB, remains to be established and represents a future direction of research.

In conclusion, our results elucidate the mechanisms by which ACTN4 regulates cancer cell invasiveness. These include regulation of focal adhesion turnover, control over protrusion dynamics and contractility via its interactions with actomyosin network, and regulation of nuclear deformation by modulation of NMM IIB expression. Given the phenotypic differences between different cancer stem cell subpopulations (None et al., 2021), it would be interesting to probe the relationship between ACTN4 expression and cancer stemness across different cancers.

MCF-7, T-47D, ZR-75-1, MDA-MB-231 and HT-1080 cells were obtained from National Center for Cell Science (NCCS, Pune, India) with MCF-7 cells cultured in MEM (Gibco, Cat # 61100-061), T-47D and ZR-75-1 cells cultured in RPMI-1640 (Gibco, Ref # 31800-022), and MDA-MB-231 and HT-1080 cells grown in high-glucose DMEM (Invitrogen, Ref # 11965084). The medium was also supplemented with 10% fetal bovine serum (FBS, HiMedia, Cat # RM9952) and with non-essential amino acids (MEM-NEAA, Gibco, Ref# 11140-050) for MCF-7 cells. Cells were maintained at 37°C at 5% CO₂ and passaged at 70–80% confluency using 0.25% trypsin-EDTA (HiMedia, Cat # TCL099). For all experiments, substrates were coated with 10 μg/cm² collagen type-I from rat tail (Sigma, Cat # C3867) overnight at 4°C.

Stable ACTN4 knockdown cell lines were generated using MISSION^® Lentiviral Transduction Particles (Cat # SHCLNV, SigmaAldrich) as per standard manufacturer's protocol. The sequences of the constructs are: 5′-CCGGCCTGTCACCAACCTGAACAATCTCGAGATTGTTCAGGTTGGTGACAGGTTTTTG-3′ (for shAC#1 cells) and 5′-CCGGGCCACACTATCGGACATCAAACTCGAGTTTGATGTCCGATAGTGTGGCTTTTTG-3′ (for shAC#2 cells). Stably transfected clones were selected by growing cells in the presence of puromycin (1.8 μg/ml for MDA-MB-231, and 1.5 μg/ml for HT-1080). Similar transfections carried out with an empty vector (Sigma, Cat # SHC001V) served as control. For fluorescence time-lapse studies, cells were transfected using transfection grade Polyethylenimine (Polysciences, Cat # 23966) or Lipofectamine 3000 (Thermo, Cat # L3000001) as per a standard protocol.

To determine the proliferation rate of control and knockdown cells, cells were sparsely seeded at equal seeding density (5000 cells/cm²) in 96-well plates and cultured for 6 days. Cells were then stained with Calcein AM (Thermo Fisher Scientific, Cat # C1430) and manually counted to obtain proliferation rate normalized to that of control cells.

For fabricating collagen gels, required amounts of rat tail collagen type-1 (Corning, Ref# 354249) stock solution was mixed with 10× PBS and complete cell culture medium (media are as above), and the pH of the solution adjusted to 7.3 using 1 M NaOH solution. The collagen solutions were then incubated in the presence or absence of cells at 37°C for 60–90 min for forming gels. For characterization of collagen gels, following snap freezing and fracturing of the hydrogels with an EM-grade blade, samples were mounted on the Cryo-unit (PP3000T, Quorum) of a JSM-7600F Cryo FEG-SEM. Samples were coated with a thin layer of platinum, and then images were obtained at desired magnification. Pore size was quantified using Fiji ImageJ.

For the gel compaction assay, 2×10⁴ cells were co-polymerized with 1.5 mg/ml rat tail collagen type-1 (Corning, Ref# 354249) on 48 well plates (200 µl/well) for 1 h at 37°C and were cultured overnight in DMEM. After 24 h of culture, collagen gels were carefully released from the edges of the wells using sharp needles. At 48 h after releasing the gels, the plate was imaged using a ChemiDoc™ imaging system (BioRad). Gel compaction was quantified by measuring reduction in gel area calculated by the expression ⁠.

For kymograph experiments, cells were sparsely seeded in collagen-coated 35 mm dishes (Greiner) and live-cell images were captured at 5 s intervals for 15 min using an inverted phase-contrast microscope (Nikon Eclipse Ti, 20× objective). Protrusion and retraction rates were analyzed from obtained images using protocols as described previously (Doyle et al., 2012). For live-cell images of myosin IIA dynamics and cofilin dynamics, cells transfected with NMM IIA–eGFP plasmid (a gift from Dr Robert Adelstein, NIH, USA) or mCherry–cofilin plasmid (Addgene plasmid #27687; deposited by Christien Merrifield) were seeded on collagen-coated glass bottom dishes and images were captured at 5 s intervals for 20 min using a 63× objective in a scanning probe confocal microscope (Zeiss, LSM 780). Kymographs were obtained by processing acquired images using Fiji ImageJ. For probing focal adhesion dynamics, cells were transected with an mCherry–paxillin plasmid (a gift from Prof. Aurnab Ghose, IISER Pune, India), seeded on collagen-coated glass bottom dishes and were imaged at 30 s intervals for 30–45 min using a 63× objective in a scanning probe confocal microscope (Zeiss, LSM 780). Focal adhesion lifetime was analyzed from acquired normalized image sequences using the focal adhesion analysis server (FAAS) as described elsewhere (Berginski and Gomez, 2013). Detection threshold in the FAAS was varied between 1.5–3.5 depending on the image signal intensity based on output from focal adhesion threshold testing.

For 2D motility studies, cells were sparsely seeded (2000 cells/cm²) in collagen-coated 48-well plates. At 12–14 h after seeding, motility videos were captured for 12 h at 15 min intervals in a live-cell imaging chamber (Tokai Hit) using an inverted microscope (Olympus IX83). Cell trajectories and speed were measured from obtained videos using the manual tracking plugin in Fiji ImageJ.

For 3D cell motility studies, cells harvested from 70% confluent cell culture dishes were mixed with pre-cursor collagen gel solutions at a density of 18,000 cells/ml, such that the final concentration of the solution was 1.2 mg/ml for MDA-MB-231 and 1.5 mg/ml for HT-1080. 180 μl of this collagen–cell mixture was dispensed in each well of glutaraldehyde functionalized 48-well cell culture plates and incubated at 37°C for gel formation. After addition of medium on top of the gels, cells were cultured for another 16–18 h inside the collagen gels prior to imaging. For MMP inhibition experiments, medium was supplemented with 15 μM GM6001. 3D motility videos were captured and analyzed the same way as described above for the 2D motility experiments. Cell-embedded hydrogels were also imaged using confocal reflection microscopy (CRM) using a 43× objective in a scanning probe confocal microscope (Zeiss, LSM 780).

For transwell migration assays, 5×10⁵ cells were seeded on the upper chamber of 24-well plate cell culture inserts containing 3 μm pores (Merck, Cat # 353096) coated with rat-tail collagen I (Sigma, Cat # C3867). For creating a gradient, the upper chamber was filled with plain medium and the lower chamber was filled with medium supplemented with 20% FBS. At 24 h (for HT 1080) to 48 h (for MDA-MB-231) after seeding, cells were fixed with 4% PFA. To count cell numbers, fixed cells were stained with DAPI (10 min). After staining, membranes were cut using a scalpel and mounted on glass slides using mounting medium. Confocal z-stack images of the fixed membranes were acquired using 20× and 63× objectives in a scanning probe confocal microscope (Zeiss, LSM 780) at identical gain and exposure settings. Images analysis and cell counting at the top and bottom of the chamber was done using ImageJ software. Translocation efficiency (η) was calculated as:

For transwell migration rescue experiments, ACTN4–eGFP-transfected knockdown cells were seeded in the transwell chambers and cells were allowed to migrate for 48 h for HT 1080 and 72 h for MDA-MB-231 after seeding. GFP-positive cells were counted at the top and bottom layer along with total cells and transfected cell ratio was calculated using the equation normalized transfected cell ratio:

Cell cortical stiffness was measured using an MFP3D Asylum AFM using an MFP3D Asylum AFM based on a protocol adapted from elsewhere (Barai et al., 2021). Sparsely seeded cells were probed using a soft pyramidal cantilever with spring constant ∼30 pN/nm (10 kHz, BL-TR400PB, Asylum Research). Cells were probed slightly away from their center to minimize the effect of nuclear stiffness, also only the first 0.5–1 µm of indentation data was fitted to get the stiffness estimates. Obtained indentation data were fitted into the Hertz model to obtain estimates of cortical cell stiffness (Barai et al., 2021; Das et al., 2017, 2019).

For immunostaining, cells were cultured on collagen-coated coverslips for 16–22 h prior to fixing with 4% paraformaldehyde. Fixed cells were permeabilized in 0.1% Triton X-100 for 8–12 min, blocked using 5% BSA for 2 h at room temperature, and then incubated with primary antibodies [anti-α-actinin 4 rabbit monoclonal antibody (Abcam, Cat # ab108198; 1:500), anti-paxillin rabbit monoclonal antibody (Abcam, Cat #ab32084; 1:500), anti-non-muscle myosin IIA rabbit polyclonal antibody (Abcam, Cat # ab75590; 1:500), anti-non-muscle myosin IIB mouse monoclonal antibody (Abcam, Cat # ab684; 1:500), anti-myosin light chain 2 (phospho-Ser19) rabbit monoclonal antibody (Cat # 3671, Cell Signaling; 1:250)] diluted in PBS overnight at 4°C. Coverslips were then washed with PBS and were incubated for 1.5 h at room temperature (RT) with appropriate secondary antibodies diluted in PBS. Finally, nuclei were stained with Hoechst 33342 (Cat# B2261, Merck) for 5 min at RT and coverslips were mounted using mounting medium. For, F-actin staining after secondary antibody incubation, Alexa Fluor 488- or Alexa Fluor 555-conjugated phalloidin (Thermo Fisher Scientific, Cat# A12379 and Cat# A12380) was added onto the coverslips for 2 h at room temperature in dark conditions. Mounted coverslips were imaged using a scanning probe confocal microscope (LSM 780, Zeiss) using a 63× objective. Images were processed and analyzed using Fiji ImageJ software. Filament number and length of was quantified from F-actin-stained images using an open-source Filamentsensor software (Eltzner et al., 2015).

Total RNA was eluted from knockdown and control cells using the RNeasy^® Plus mini kit (Qiagen, Ref # 74134). 2 µg of total RNA was reverse transcribed using a High Capacity cDNA Reverse Transcription Kit (AppliedBiosystems, Ref. # 4368814) and was amplified using PowerUp^TM SYBR Green Master Mix (AppliedBiosystems, Ref. # A25742) in a QuantStudio 5 (AppliedBiosystems). PCR data were analyzed using comparative Ct method and cyclophilin A expression was used to normalize gene expression data. Primers used for RT-PCR are as follows: ACTN1-Forward, 5′-CAAAGATTGATCAGCTGGAG-3′; ACTN1-Reverse, 5′-CTCTACCTCATTGATGGTCC-3′; ACTN4-Forward, 5′-AGTATGACAAGCTGAGGAAG-3′; ACTN4-Reverse, 5′-CTGAAAAGGCATGGTAGAAG-3′; NMM IIA-Forward; 5′-GTGAAGAATGACAACTCCTC-3′; NMM IIA-Reverse, 5′-GATAAGTCTCTCAATGTTGGCTC-3′; NMM IIB-Forward, 5′-CACTGAGAAGAAGCTGAAAG-3′; NMM IIB-Reverse, 5′-TCTGCTCTTTATACTGGTCC-3′, Cyclophilin A-Forward, 5′-TGGGCCGCGTCTCCTTTGA-3′; Cyclophilin A-Reverse, 5′-GGACTTGCCACCAGTGCCATTA-3′.

Whole-cell lysates were prepared using RIPA buffer (Sigma, Cat # R0278) containing protease inhibitor cocktail (Sigma, Cat # P8340) and phosphatase inhibitor cocktail (Sigma, Cat # P0044), were then centrifuged at 12,000 g for 10 min and protein concentration of the collected supernatant was determined using Bradford reagent. SDS-PAGE was performed with an equal amount (20–30 µg) of protein per lane and then were transferred onto a nitrocellulose membrane (PALL Life Sciences, Cat # 66485). Following transfer, the membranes were blocked with 5% BSA prepared in TBST (10 mM Tris-HCl, pH 8.0 containing 150 mM NaCl and 0.1% Tween 20) for 1 h at room temperature and were then incubated overnight at 4°C with the following primary antibodies diluted in TBST: anti-α-actinin-4 rabbit monoclonal antibody (Abcam, Cat # ab108198; 1:1200), anti-vimentin mouse monoclonal antibody (Abcam, Cat # ab8978; 1:1000), anti-integrin β1 rabbit monoclonal antibody (Abcam, Cat # ab155145; 1:1000), anti-GAPDH rabbit polyclonal antibody (Abcam, Cat # ab9485; 1:2000), anti-lamin A/C (phosphoS392) rabbit polyclonal antibody (Abcam, Cat # ab58528; 1:1500), anti-lamin A/C (phosphoS22) rabbit polyclonal antibody (Cat # 2026, Cell Signaling Technology; 1:1500), anti-lamin A/C mouse monoclonal antibody (Abcam, Cat # ab8984; 1:1000), anti-non-muscle myosin IIA rabbit polyclonal antibody (Abcam, Cat #ab75590; 1:1000), anti-non-muscle myosin IIB mouse monoclonal antibody (Abcam, Cat # ab684; 1:2000). Blots were then washed with TBST and were incubated with HRP-conjugated anti-rabbit IgG (Invitrogen) or HRP-conjugated anti-mouse IgG (Invitrogen) secondary antibodies for 1.5 h at room temperature. Finally, membranes were washed three times with TBST and were developed using a chemiluminescent ECL kit (Pierce, Cat # 32106) and images were acquired using a ChemiDoc™ imaging system (BioRad).

Co-IP was performed using Protein G Immunoprecipitation Kit (Merck, Cat # IP50) as per the manufacturer's protocol. Briefly, whole-cell lysates (250 µg) were incubated at 4°C for 12 h with anti-α-actinin-4 antibody (Abcam, Cat # ab108198; 2.5 µg) or anti-non-muscle myosin IIA antibody (Abcam, Cat #ab75590; 2 µg). 30 μl slurry of protein G-agarose beads were added to the immunocomplexes in the IP spin column and incubated for 6 h at 4°C. After 6–7 times vigorous washing in the spin column using 1× IP buffer, pellets were eluted from the column after heating for 5 min at 95°C with 1× Laemmli buffer. Eluted samples along with 10 µg whole-cell lysate as a control were then subjected to SDS-PAGE and western blot analysis using anti-non-muscle myosin IIA antibody, anti-α-actinin 4 antibody and anti-GAPDH antibody (Abcam, Cat # ab9485; 1:2000) as described above.

Duolink^® PLA Red Starter Kit (Merck, Cat # DUO92101), which allows for endogenous detection of protein interactions, was used to detect ACTN4–NMM IIA interaction. The kit was used as per the standard manufacturer's protocol using rabbit anti-α-actinin 4 antibody (Abcam, Cat# ab108198; 1:500) and anti-mouse non-muscle myosin IIA antibody (Novus, Cat# H00004627-M06). The PLA signals were visible as red fluorescent spots when imaged using a confocal microscope (Zeiss, LSM 780).

Data distribution was tested using the Kolmogorov–Smirnov normality test. Based on the outcome, either a parametric or nonparametric statistical test was performed. For parametric data, statistical analysis was performed using one-way ANOVA and Tukey's test was used to compare the means. A Mann–Whitney test was performed for non-parametric data. Statistical analysis was performed using Origin 9.1 (OriginLab^® Corporation), with P<0.05 considered to be statistically significant.

NMM IIA was docked to either a 6-monomer F-actin filament or ACTN4-bound F-actin filament using the ClusPro protein docking server (Kozakov et al., 2017). For these studies, the predicted full-length ACTN4 (https://alphafold.ebi.ac.uk/entry/O43707) and NMM IIA (https://alphafold.ebi.ac.uk/entry/P35579) structures were obtained from AlphaFold protein structure database (Jumper et al., 2021). The structure of the 6-monomer F-actin filament was generated using Discovery Studio Visualizer based on a cryo-electron microscopy structure of the α-actinin CH1 domain bound to F-actin (PDB: 3LUE). ACTN4-bound F-actin was generated based on a full-length F-actin alignment to CH1-bound F-actin using PyMOL. Interaction sites were identified based on the Cryo-EM structure of a human cytoplasmic actomyosin complex (PDB: 5JLH). Predicted models were analyzed and images were generated using PyMOL.

For studying the importance of ACTN4-mediated cell protrusive forces in regulating confined cell migration, a plane strain finite element (FE) model of confined migration was created in ABAQUS using an explicit formulation for resolving large deformations, similar to that found in our recent publication (Mukherjee et al., 2020). In this model, pore entry of a 10 µm diameter cell with a 5 µm diameter nucleus was simulated through a deformable matrix for varying pore sizes (ϕ={3, 5, 8} µm) (Fig. 6A). The cell-matrix interface was assumed to be frictionless. The components of the cell, including the cell cortex (0.5 µm thick) and the cytoplasm were modeled as Kelvin–Voigt viscoelastic elements with the viscoelastic character of each component represented in the form of normalized creep compliance (Fig. S7B). The nucleus was modeled as an elastic solid encircled with a viscoelastic 50 nm thick membrane mimicking the Lamin-rich nuclear periphery (Hsu and Kao, 2013). The surrounding tissue was also assumed to be viscoelastic in nature. A Poisson's ratio of ν=0.3, which is typical of compressible biomaterials (Hsu and Kao, 2013), was chosen for all the viscoelastic elements and is consistent with experimental observations of cell migration through microchannels wherein cells get polarized in the direction of migration, but the accompanying lateral deformations are negligible (Lu et al., 2012; Pickup et al., 2014). All elastic material properties are listed in Table S1.

In our model, pore entry was mediated by active protrusive forces exerted on 250 individual nodes at the cell front such that the magnitude of the maximum force generated by the cell (F_Pmax) remained in the physiologically relevant range (Rabodzey et al., 2008) (Fig. S7C). F_Pmax was chosen to vary between 1.25 nN/µm and 6.25 nN/µm to mimic different levels of ACTN4. In line with our experimental observations, the cortical stiffness (E_cortex) was also changed accordingly. The simulation was stopped once the nucleus entered the pore completely.

Different sections of the cell and the tissue were meshed such that a smaller element size was used at regions expected to undergo large deformations and/or coming in contact with the matrix. A total of 30,551 bilinear plane strain CPE4R elements were used in the model with a minimum element dimension of 5 nm and a maximum of 20 µm. Mesh gradation of the entire system is shown in Fig. 6A. The nuclear membrane, 50 nm in thickness, had ten elements in the through-thickness direction to mitigate the effects of excessive artificial bending stiffness.

The authors thank Dr Alan Wells (Univ. of Pittsburg) for sharing the eGFP–ACTN4 plasmid and Prof. Aurnab Ghose (IISER Pune) for sharing the mCherry–Paxillin plasmid, and Dr Robert Adelstein (NIH, USA) for the GFP–NMM IIA plasmid. We acknowledge IIT Bombay for providing Bio-AFM, Cryo FEG-SEM and Confocal Microscopy facilities.

The peer review history is available online at https://journals.biologists.com/jcs/article-lookup/doi/10.1242/jcs.258581.