Contact guidance requires spatial control of leading-edge protrusion
In vivo, geometric cues from the extracellular matrix (ECM) are critical for the regulation of cell shape, adhesion, and migration. During contact guidance, the fibrillar architecture of the ECM promotes an elongated cell shape and migration along the fibrils. The subcellular mechanisms by which cells sense ECM geometry and translate it into changes in shape and migration direction are not understood. Here we pattern linear fibronectin features to mimic fibrillar ECM and elucidate the mechanisms of contact guidance. By systematically varying patterned line spacing, we show that a 2-μm spacing is sufficient to promote cell shape elongation and migration parallel to the ECM, or contact guidance. As line spacing is increased, contact guidance increases without affecting migration speed. To elucidate the subcellular mechanisms of contact guidance, we analyze quantitatively protrusion dynamics and find that the structured ECM orients cellular protrusions parallel to the ECM. This spatial organization of protrusion relies on myosin II contractility, and feedback between adhesion and Rac-mediated protrusive activity, such that we find Arp2/3 inhibition can promote contact guidance. Together our data support a model for contact guidance in which the ECM enforces spatial constraints on the lamellipodia that result in cell shape elongation and enforce migration direction.
Cell migration plays a central role in several developmental, physiological, and pathological processes. During development, directed migration is required for diverse morphogenetic processes conserved among organisms, ranging from branching morphogenesis of kidney and breast tissue, to migration of neural crest cells out of the tube (Keller, 2005; Vasilyev et al., 2009; Brownfield et al., 2013). In adults, leukocytes are required to translocate to areas of insult to mediate phagocytic and immune function (Luster et al., 2005). Migration of fibroblasts and vascular endothelial cells is essential for wound healing (Gillitzer and Goebeler, 2001). In cancer metastasis, multicellular masses invade their extracellular matrix (ECM) and migrate into the bloodstream (Friedl and Gilmour, 2009). Migration of single cells has been studied extensively and is relatively well understood. Despite its complexity, cell migration can be thought of as consisting of four discrete steps: protrusion of the cell’s leading cell edge, adhesion to the ECM, generation of traction stresses against adhesions, and release of rear adhesions and cell body contraction (Gardel et al., 2010). To get productive advancement, cells must first polarize to form a leading front and a back. The ability of cells to polarize is controlled by numerous external and internal cues. Internally, membrane tension can restrict competing growing leading edges (Houk et al., 2012), and regions of high Rho activity can inhibit cell protrusion (Wang et al., 2003), whereas externally, physical characteristics of the ECM such as rigidity, density, and geometry can induce cell polarity, including directed migration and elongation of cell shape (Petrie et al., 2009; Charras and Sahai, 2014).
Decades of work have shown that changes in geometry, or spatial arrangement, of adhesive cues provided by the ECM can align cell shape and guide cell motility (Weiss and Garber, 1952; Dunn and Ebendal, 1978; Clark et al., 1990; den Braber et al., 1996a; Teixeira et al., 2003; Provenzano et al., 2008; Xia et al., 2008). During this process, termed contact guidance, the geometry of the ECM enforces changes to cell morphology and migration direction (Petrie et al., 2009; Reig et al., 2014). Classical studies in vivo and in vitro have shown that cells interacting with fibrillar ECM networks exhibit contact guidance (Weiss and Garber, 1952; Dunn and Heath, 1976; Dunn and Ebendal, 1978). For instance, during amphibian gastrulation, aligned ECM fibrils facilitate mesodermal cell migration toward the animal pole (Nakatsuji et al., 1982), whereas cells interacting with fibrin clots reorient the ECM, which in turn alters their morphology and migration (Weiss and Garber, 1952).
Micropatterning has been a powerful tool to demonstrate that changes in the geometrical presentation of ligands alone, such as area, shape, and density, can regulate cell shape and migration (Clark et al., 1990; den Braber et al., 1996a; Teixeira et al., 2003; Provenzano et al., 2008; Xia et al., 2008; Doyle et al., 2009; Yoon et al., 2012; Londono et al., 2014). Significantly less explored are the subcellular mechanisms used by cells to sense changes in ECM geometrical parameters. In patterned environments, lamellipodia and fillipodia extend parallel to the ECM (Teixeira et al., 2003, 2006). Extension of lamellipodia parallel to the ECM has been correlated with localized Rac activity at adhesions (Xia et al., 2008) and actin polymerization parallel to the ECM (Sun et al., 2015), whereas inhibition of myosin II contractility has been shown to be important to maintain cell shapes in response to ECM architecture (Frey et al., 2006; Spedden et al., 2014). In these studies, along with ECM orientation, density of available ECM (Xia et al., 2008) and surface topography (Frey et al., 2006; Spedden et al., 2014; Sun et al., 2015) were varied, making it impossible to discern effects of ECM alignment alone. Given that the ECM in vivo typically comprises a network of fibers with micrometer-scale structure (Nakatsuji et al., 1982; Condeelis and Segall, 2003; Baker and Chen, 2012), we sought to understand the mechanism by which interfibrillar spacing controls the tendency of cells to align and move along fibrils.
Contact guidance can be tuned by micrometer-scale changes in ECM geometry
To mimic fibrillar geometry and isolate the effects of variations in interfibrillar spacing, we micropatterned parallel lines of fibronectin of a constant width (2 µm) with interline spacing varying from 0 (uniform) to 10 µm (Oakes et al., 2014). These dimensions were chosen such that NIH 3T3 fibroblasts were able to spread on multiple lines, and the line width was large enough for typical focal adhesion formation (Lehnert et al., 2004; Doyle et al., 2009). Fibronectin was patterned onto polyacrylamide gels with a stiffness that allowed for optimal fibroblast migration (Lo et al., 2000). Immunofluorescence of fibronectin, paxillin, and actin confirmed that focal adhesions were excluded from unpatterned regions and formed on fibronectin stripes, allowing cells to attach to several pattern features (Figure 1, A and B).
Fibroblasts plated on uniformly coated substrates (0-µm spacing) acquired a characteristic polarized morphology with no preferred directional orientation (Mogilner and Keren, 2009). On all line patterns, cells elongated and became preferentially aligned to the ECM (Figure 1B). To quantify these changes in shape and orientation, we fitted an ellipse to each cell and characterized the elongation as the ratio between the long and short axes, with the alignment to the axis defined by the ECM lines with the parameter P = cos2θ, where θ is the angle between a line parallel to the ECM and the long axis of the cell ellipse. Cell elongation increased linearly with line spacing (Figure 1D), whereas cell spread area decreased (Supplemental Figure S1A). Orientation increased sharply from 0- to 2-µm spacing, with a mean orientation with respect to the ECM lines of 0.81 out of a maximum possible value of 1. The orientation between 2 and 10 µm then slowly increased until it reached a mean value of 0.98, indicating almost complete alignment with the ECM (Figure 1E). Thus increased fibril spacing promoted an elongated cell morphology, with the long axis parallel to the line direction.
During cell migration, the long axis typically defines the axis of front–back polarity and corresponds to the direction of migration (Lauffenburger and Horwitz, 1996; Petrie et al., 2009). Thus we next sought to determine how these changes in ECM presentation affected the speed and direction of spontaneous cell migration. Figure 1C shows rose plots of the trajectories of ∼100 cells during a period of 10 h. Strikingly, the instantaneous migration speed was unaffected over this range of line spacings (Figure 1F), but the cell trajectories became oriented parallel to the ECM with increased line spacing (Figure 1C and Supplemental Movie S1). To quantify this effect, we defined a binary guidance (Londono et al., 2014) that measured the orientation of a cell’s trajectory relative to the ECM. Briefly, we calculated the angle θ between the displacement vector of a cell and the ECM and defined binary guidance, g, such that g(t) = 0 if θ > 25 and g(t) = 0 if θ ≤ 25. When we averaged this measurement over several time intervals τ, we obtained a guidance parameter which represents accurately the direction of a cell’s trajectory over different time scales (Supplemental Figure S1B). The value of G at τ = 300 min is reported because it is the relevant time scale for the experiments performed. The guidance parameter demonstrates that migration direction is increasingly parallel to the ECM as a function of line spacing, having its sharpest increase from 2 to 3 µm (Figure 1G). Consistent with the changes observed in cell migration and cell shape, cells oriented their traction stresses and migration direction parallel to the ECM on line patterns (Supplemental Figure S1).
Taken together, these results demonstrate that micrometer-scale variations in fibril-like spacing from 2 to 3 µm can tune cell shape and bias the direction of cell migration parallel to the ECM. This is consistent with previous studies, which found that micrometer-scale changes in pattern spacing can induce cell shape alignment (Clark et al., 1990; den Braber et al., 1996a; Teixeira et al., 2003; Xia et al., 2008). However, in these studies, along with line spacing, size of adhesive regions (Xia et al., 2008) or surface topography was varied (Clark et al., 1990; den Braber et al., 1996b; Teixeira et al., 2003), making it impossible to discern the effects of line spacing alone.
Leading-edge protrusions are oriented parallel to the ECM on line patterns via myosin II contractility and ECM feedback
Cell migration is controlled by the spatiotemporal organization of leading-edge protrusion, adhesion, and cell retraction (Gardel et al., 2010). To investigate how protrusion dynamics were altered during contact guidance, we followed shape changes in cells transfected with a fluorescent membrane marker (green fluorescent protein [GFP]–stargazin) that facilitated the identification of the cell edge (Figure 2A, Supplemental Figure S2A, and Supplemental Movie S2). From the cell contours, we identified the area and orientation of protrusive regions (Figure 2B). We found that the area of individual protrusions was unaffected by changes in line spacing (Figure 2C). However, as the line spacing increased, protrusions became oriented with respect to the ECM lines (Figure 2D). This result suggested that changes in ECM line spacing most significantly affected the spatial organization of protrusions rather than their overall size. This is consistent with our observation that contact guidance is promoted even at line spacings of 2–3 µm that would not prevent the stabilization of lamellipodial protrusions.
The spatial organization of lamellipodial activity is controlled by numerous intracellular and extracellular cues (Raftopoulou and Hall, 2004; Charras and Sahai, 2014). Most relevant to this work are the well-established feedbacks between cell shape (Parker et al., 2002; James et al., 2008; Gabella et al., 2014; Elliott et al., 2015) and cell–matrix adhesion to local protrusive activity (Nayal et al., 2006; Xia et al., 2008). The feedback between cell shape and protrusion regulation is mediated by myosin II contractility. Along the long, nonadherent edges of cells, myosin II–mediated contractility suppresses local protrusive activity (Elliott et al., 2015). At focal adhesion sites, positive feedback between cell–matrix adhesion and protrusive activity is achieved by activation of Rac through β-pix–mediated focal adhesion signaling (Nayal et al., 2006).
To explore whether the elongated shape plays an important role in the spatial organization of protrusion we observe, we analyzed spreading and protrusion dynamics in the initial stages of cell spreading. On uniformly coated substrates, cells undergo rapid and Rac1-mediated isotropic spreading after initial attachment (Price et al., 1998; Wolfenson et al., 2014). By contrast, cells spread anisotropically on line patterns (Figure 2E, left, and Supplemental Movie S3, top) and quickly oriented along linear features within 4 min after initial contact (Figure 2G). After the initial orientation, cell elongation continued to increase (Figure 2F) even after cells had reached their maximum spread area (Supplemental Figure S2B). The rapid orientation of cell shape on ECM attachment suggests that positive feedback with the ECM may be sufficient to guide protrusion dynamics even in the absence of myosin II–mediated regulation. To determine the role of myosin II activity in the ability of cells to orient along ECM lines during initial spreading, we observed spreading dynamics in cells treated with 20 µM of the Rho kinase (ROCK) inhibitor Y-27632. Surprisingly, ROCK-inhibited cells failed to elongate during spreading on all line spacings (Figures 2, E, right, and F, and Supplemental Movie S3, bottom). Furthermore, cells had severe defects orienting parallel to the ECM (Figure 2G), achieving orientation only at 10-µm spacing at later times during spreading, showing that ROCK-mediated contractility is necessary to direct protrusions parallel to the ECM, even in the earliest stages of cell–ECM attachment.
These results demonstrate that structured ECM regulates protrusive activity spatially to enforce cell shape polarization along the axis of the ECM lines. The altered local density of ECM did not affect the overall protrusion size but biased their formation preferentially along the ECM axis even at spacings of 2–3 µm. Because lamellipodia are large enough to span those gap sizes (Machacek et al., 2009), this suggests that contact guidance can occur even when the size of the protrusions enables them to span line spacings. However, because protrusion area remains constant across spacings (Figure 2C), it is possible that limitations on the size of protrusions can bias their direction, particularly at spacings >3 µm. We find that spatial control of protrusive activity is regulated from the earliest stages of cell spreading. ROCK-mediated contractility is necessary to achieve cell shape polarization during spreading, suggesting that it plays an important role in contact guidance, which is consistent with its known role providing cytoskeletal coherence during spreading (Cai et al., 2010).
ROCK-mediated contractility is necessary for contact guidance
Because ROCK-mediated contractility was required to polarize cells during spreading, we next explored its role in contact guidance. We first analyzed the shape of fully spread cells after treatment with ROCK inhibitor for 45 min (Figure 3A, top). Consistent with the characteristic morphology of ROCK-inhibited cells, cells often had multiple protrusive fronts and a tail retraction defect (Figure 3A, top; Omelchenko et al., 2002). In contrast to control cells, ROCK-inhibited cells did not elongate further when placed on line patterns (Figure 3E and Supplemental Figure S3A, top). However, ROCK-inhibited cells exhibited a similar tendency to orient along the line axis to control cells (Figure 3, A, top, and F). Despite this orientation, migration directionality was severely affected in ROCK-inhibited cells (Figure 3A, bottom, Supplemental Figure S3A, bottom, and Supplemental Movie S4). Compared to control cells, the guidance was reduced by 50% (Figure 3, G and J). Thus inhibition of ROCK-mediated contractility results in a decoupling between cell orientation and guidance.
To determine the mechanisms causing the migration guidance defect in ROCK-inhibited cells, we analyzed protrusions using the method described earlier. Consistent with known effects of ROCK inhibition (Omelchenko et al., 2002), cells had disorganized protrusions and poorly defined leading-edge fronts (Figure 4A, Supplemental Figure S4A, and Supplemental Movies S8 and S9). Protrusion area doubled relative to controls for spacings <10 µm (Figure 4, E and G). More strikingly, protrusions failed to organize parallel to the ECM, as shown by a sharp decrease in orientation for line spacings <10 µm (Figure 4F). Compared to control cells, the orientation of protrusions decreased by 20% (Figure 4H).
Together these results show that ROCK-mediated contractility is necessary to organize protrusions spatially in response to structured ECM, promoting contact guidance. Previous work suggested that inhibition of myosin II contractility reduced the ability of cells to orient parallel to the ECM surface microtopography and nanotopography (Frey et al., 2006; Spedden et al., 2014). Our results demonstrate that the mechanism is likely through the important role of ROCK-mediated contractility in spatial organization of protrusions (Lo et al., 2004).
Perturbations to protrusive activity via misregulation of Rac 1 impair contact guidance
Spatial regulation of Rac activity controls the formation, growth, and size of leading-edge protrusions (Nobes and Hall, 1999; Ridley et al., 2003; Pankov et al., 2005; Machacek et al., 2009). High levels of Rac activity were previously shown to impede directed migration by promoting the formation of several lamellipodial fronts (Pankov et al., 2005). Therefore, to promote global activation of Rac and disrupt the spatial regulation of lamellipodia formation, we expressed the constitutively active form RacQ61L. RacQ61L cells were rounder (Figure 3B top, and Supplemental Figure S3D), as shown by the 50% decrease in elongation for all line spacings relative to wild-type (WT) cells (Figure 3H). Elongation values in these cells were >1, allowing us to define a long axis to measure cell shape orientation (Supplemental Figure S3D). Intriguingly, orientation values in RacQ61L cells were comparable to those in WT cells (Supplemental Figure S3E), as shown by the ratio between the values being close to 1 (Figure 3I). In contrast, migration trajectories were no longer parallel to the ECM (Figure 3B, bottom, and Supplemental Movie S5), except at 10-µm line spacing, where cells exhibited limited guidance (Supplemental Figure S3F). The defect in guidance was more pronounced than for ROCK-inhibited cells, because for all line spacings, guidance in RacQ61L cells was reduced by at least 50% (Figure 3J).
To determine how protrusive activity was altered in cells expressing RacQ61L, we next measured area and localization of leading-edge protrusion. First, we noticed that cells exhibited protrusions around the entire cell body (Figure 4B and Supplemental Movie S10). Measurements showed that protrusion areas were comparable to those in WT cells (Figure 4G and Supplemental Figure S4D). By contrast, protrusions failed to orient parallel to the ECM as a function of line spacing (Supplemental Figure S4E), decreasing in orientation by 30% relative to WT cells (Figure 4H). This trend of protrusion orientation mimics the trend observed for guidance (compare Figures 3J and 4H), suggesting that the ability of cells to migrate parallel to the ECM relies on biasing protrusive activity in the direction parallel to the lines.
Another major regulator of protrusive activity relevant to our context is the positive feedback between sites of new adhesion and protrusion formation, controlled by p21-activated kinase (PAK)–mediated changes in paxillin phosphorylation, which recruits the Rac guanine nucleotide exchange factor (GEF) β-pix (Nayal et al., 2006). Knockdown of β-pix has been linked to increased lamellipodia protrusion and defects in directed migration (Kutys and Yamada, 2014; Omelchenko et al., 2014). To explore how disruption to this feedback affected contact guidance, we knocked down β-pix to 30% of its endogenous levels (Supplemental Figure S3M). For cells on lines, β-pix knockdown (KD) led to a slight reduction in elongation for line spacings >3 µm (Figure 3C, top, and Supplemental Figure S3B, top), reaching only 70% of the nontargeting-control (NT) cell values (Figure 3H and Supplemental Figure S3G).
Similar to previously analyzed cells, β-pix KD had minimal defects in cell orientation parallel to the ECM (Figure 3I and Supplemental Figure S3H). Strikingly, migration guidance was reduced in these cells by ∼50% compared with the NT-control cells (Figure 3J, Supplemental Figure S3I, and Supplemental Movie S6). Together these results show that β-pix feedback between adhesion and protrusion initiation is not essential for cell shape anisotropies during contact guidance; however, it plays an important role in directing migration in the direction of the ECM. Because β-pix is required for the positive feedback between adhesion and protrusion, we expected β-pix KD cells to mislocalize protrusions similarly to RacQ61L cells. Surprisingly, cell protrusions occurred mostly along the ECM axis (Figure 4C, Supplemental Figure S4B, and Supplemental Movies S11 and S12). Quantification of protrusion orientation revealed that protrusions localized parallel to the ECM to levels comparable to that for NT control cells (Figure 4H and Supplemental Figure S4G). In contrast with RacQ61L cells, β-pix KD cells had large protrusions that fluctuated greatly in size, as evidenced by the twofold increase in protrusion area and the large fluctuations in this measurement (Figure 4G and Supplemental Figure S4F).
Taken together, these results show that spatial regulation of protrusive activity is required for contact guidance. Of importance, in each of these phenotypes, the guidance can be significantly affected without much change to alignment of the cell long axis in the directions of the lines. Our β-pix KD data suggest that in addition to localization of protrusions, the size of protrusions can regulate the directed migration for cells on patterns. This supports our previous hypothesis in which the size of protrusions can limit the ability of cells to migrate perpendicular to the ECM. To further test this idea, we decided to explore in detail whether changes to protrusion size could affect contact guidance.
Arp2/3 inhibition increases contact guidance
The Arp2/3 complex is an essential component of lamellipodial protrusions at the leading edge (Wu et al., 2012; Beckham et al., 2014). Previous work showed that lamellipodia-driven migration plays an important role in haptotaxis (Wu et al., 2012; King et al., 2016), suggesting that it also might also impede contact guidance. To explore this, we treated cells with 100 µM Arp2/3 inhibitor CK-666. On first inspection, Arp 2/3-inhibited cells appeared to be more elongated and polarized in the direction of the ECM (Figure 3D, top, and Supplemental Figure S3C, top). Quantification confirmed that these cells had increased elongation compared with cells treated with the control molecule CK-689, particularly at line spacings <5 µm (Figure 3H and Supplemental Figure S3J), whereas cell orientation was only slightly higher than for controls (Figure 3I and Supplemental Figure S3K). Of greater significance, we observed higher values for guidance at all line spacings, with a twofold increase at spacings <5 µm compared with controls (Figure 3J, Supplemental Figure S3L, and Supplemental Movie S7). These measurements provide the first observation of an increase in guidance compared with control cells.
Our previous results show that contact guidance depends inversely on protrusion size and directly on localization of protrusions parallel to the ECM. We therefore expected Arp 2/3-inhibited cells to have reduced protrusion size and/or highly localized protrusions. Indeed, cell protrusions appeared to be smaller and directed almost exclusively in the direction of the ECM (Figure 4D, Supplemental Figure S4C, and Supplemental Movies S13 and S14). Quantitative analysis confirmed that protrusions in Arp2/3-inhibited cells decreased by 40% in area compared with cells treated with the control molecule (Figure 4G and Supplemental Figure S4H). Remarkably, orientation of protrusions followed the same trend as migration guidance, increasing 1.2-fold for cells on line spacings <5 µm (Figure 4H and Supplemental Figure S4I). However, it is possible that during contact guidance, the spatial constraints of adhesive cues are sufficient to organize adhesion and direct protrusions.
Together these results further demonstrate that both the area and orientation of lamellipodial protrusions contribute to contact guidance. The poorest contact guidance occurs in cells with reduced orientation, such as what occurs in ROCK-inhibited or RacQ61L-expressing cells (Figure 4, G and H). Alternatively, contact guidance can be improved over the WT condition by reduced overall lamellipodial size and improved orientation.
The ability of cells to change shape and direct their migration in response to ECM geometric cues has been characterized for decades and remains an area of active research (Dunn and Ebendal, 1978; Nakatsuji et al., 1982; Dickinson et al., 1994; den Braber et al., 1996a; Teixeira et al., 2003; Xia et al., 2008). Previous studies characterized extensively the ability of a variety of cells to respond to changes in ECM geometry at scales ranging from nanometers (Teixeira et al., 2006; Loesberg et al., 2007; Jang et al., 2010) to hundreds of micrometers (Vedula et al., 2012, 2014). Despite numerous studies that pointed that the ability of cells to sense their underlying matrix is ubiquitous across cell types and occurs at different scales, few studies addressed the mechanisms of contact guidance. Efforts focused on understanding the mechanisms of contact guidance in the extreme case in which cells are confined to a single line of ECM. In this case, cells move 1.5 times faster than in a classical two-dimensional environment due to changes in lamellipodial dynamics and increased mechanical coupling between adhesions and the protrusive machinery (Doyle et al., 2009, 2012). Although the effects of different degrees of confinement on cell migration can be studied by varying the width of the ECM line, migration direction is trivially parallel to the ECM. Therefore, to understand the mechanisms by which cells adjust their direction of migration in response to changes in ECM geometry, we patterned microenvironments in which the ECM line width was constant. By patterning ECM architectures with parallel lines of a fixed width that impose no limitations on adhesion formation, we show that in the absence of confinement, cells are capable of directing migration parallel to the ECM.
Studies on cell-derived matrices and ECM protein hydrogels showed that cells orient their motility and shape in response to anisotropies in the ECM in the absence of confinement (Provenzano et al., 2008; Kutys and Yamada, 2014). These experimental systems resemble more closely the in vivo microenvironment but have the disadvantage that it is challenging to control the degree of anisotropy of the ECM with subcellular resolution. In these cases, only binary comparisons can be made (aligned ECM vs. not aligned ECM), making it impossible to discern whether cells are sensitive to a range of anisotropies in the ECM or guidance is just a binary phenomenon. By varying the spacing between lines, we are able to control the degree of anisotropy in the ECM. Thus our study addresses a gap in the literature, showing that variations in anisotropy of the ECM at the micrometer scale can influence migration in a continuous way. At the same time, we show that by changing the fibril-like spacing alone, cells can tune their guidance behavior, in contrast with previous studies, in which surface topography and fibril-like spacing were changed simultaneously (den Braber et al., 1996a,b, 1998; Flemming et al., 1999; van Kooten and von Recum, 1999; Walboomers et al., 1999a, b; Matsuzaka et al., 2000; Teixeira et al., 2003; Provenzano et al., 2008).
Our work provides the foundation for a mechanistic understanding of contact guidance at the micrometer scale. Previous work on structured ECM showed that Rac was activated at sites of adhesion formation and consequently stimulated protrusion formation at those sites (Xia et al., 2008), whereas myosin II activity was important for the cell shape changes that occur in response to structured ECM (Frey et al., 2006; Spedden et al., 2014). However, by manipulating the main feedbacks that regulate cell migration, we show that defects in contact guidance stem from defects in the spatial control of protrusive activity. Our quantitative analysis enabled the characterization of contact guidance, allowing comparison of cellular phenotypes across different perturbations. Comparison across experimental conditions shows that perturbations by which cells lose their ability to migrate parallel to the ECM fail to localize protrusions (compare Figures 3J and 4H) or have protrusions with increased area (compare Figures 3J and 4G). By contrast, Arp 2/3 inhibition improves contact guidance by a decreasing the protrusion area (compare Figures 3J and 4G) and improving localization of protrusions along the lines. Collectively, these results show that contact guidance relies on spatial control of leading-edge protrusion. The defects we observe in migration guidance can be explained by changes in two distinct parameters of protrusion: orientation and size. Specifically, contact guidance is controlled by each of these parameters; disruption of either one is sufficient to produce loss of contact guidance. These two parameters are tuned subcellularly by positive feedback between adhesion formation and protrusion initiation and negative feedback between myosin II–dependent elongated shapes and formation of protrusions perpendicular to the ECM. For spacings ≤5 µm, these two feedbacks are required to enforce elongated cell shapes and migration to the ECM. As spacing is increased, the physical size of protrusions limits the ability of protrusions to be stabilized perpendicular to the ECM, further enhancing contact guidance (Figure 5A). The spacing at which this transition occurs is most likely cell-type dependent because variations in lamellipodia size that are cell-type dependent will influence this. Thus we envision that contact guidance is a function of protrusion area and orientation, where these two parameters can be regulated through different cellular pathways (Figure 5B). Future work needs to investigate other cellular pathways that regulate the efficiency of contact guidance that may arise from focal adhesion signaling and protrusion dynamics. In addition, the connection between contact guidance at the micrometer and nanometer scales remains to be explored. Although there is little mechanistic understanding of contact guidance at the nanometer scale, there is evidence that focal adhesion and protrusion orientation correlate with the ability of cells to align parallel to the underlying matrix (Teixeira et al., 2003, 2006; Jang et al., 2010). At this scale, other molecular feedbacks might become important for ECM sensing, such as local curvature of cell shape and adhesion signaling driven by focal adhesion kinase (Houk et al., 2012; Teo et al., 2013).
MATERIALS AND METHODS
Cell culture and reagents
NIH 3T3 fibroblasts (American Type Culture Collection, Manassas, VA) were cultured in DMEM (Mediatech, Herndon, VA) and supplemented with 10% fetal bovine serum (HyClone; Thermo Fisher Scientific, Hampton, NH), 2 mM l-glutamine (Invitrogen, Carlsbad, CA), and penicillin-streptomycin (Invitrogen). The ARP2/3 inhibitor CK-666 and control compound CK-689 were purchased from Calbiochem. Y-27632 was purchased from Sigma-Aldrich. Cells were plated for 4 h and then incubated for 3 h in CK-666 or 45 min in Y-27632 before imaging or fixation at the concentration indicated for all experiments. Cells were transiently transfected with plasmid DNA constructs encoding GFP-stargazin (a gift from A. Karginov, University of Illinois at Chicago, Chicago, IL) and RacQ61L (a gift from the Gary Borisy laboratory, Northwestern University). Transfections were performed using a Neon electroporation system (Invitrogen). Short hairpin RNA (shRNA) construct sets for mouse β-pix (GIPZ Arhgef7 shRNA set) were purchased from Dharmacon through Thermo Fisher Scientific. Transfection of constructs was performed using a Neon electroporation system (Invitrogen). Puromycin and fluorescence selections of expressing cells were performed per the manufacturer’s protocols. Knockdown efficiency was determined by Western blotting (Supplemental Figure S3), and a stable cell line was maintained for the V3LMM_521424 clone.
Western blot analysis
For Western blotting, cells were lysed in radioimmunoprecipitation assay buffer (50 mM Tris, 150 mM NaCl, 0.1 SDS, 0.5 Na deoxycholate, 1 Triton X-100, 1 mM phenylmethylsulfonyl fluoride, and protease inhibitor). Lysates were separated by SDS–PAGE gel and electrotransferred to a nitrocellulose membrane.
Blots were blocked in Odyssey blocking buffer and incubated with primary antibody (rabbit anti–β-pix; 07-1450-I, EMD Millipore; rabbit anti–glyceraldehyde-3-phosphate dehydrogenase; G9545, Sigma-Aldrich) overnight at 4°C. Blots were incubated in secondary antibody donkey anti–rabbit 680LT (C50224-04; Odyssey ) for 1 h at room temperature. Blots were scanned using the Odyssey Fc imaging system and analyzed using the gel analysis tool in ImageJ (1.51f; National Institutes of Health).
Cells were rinsed in warm cytoskeleton buffer (10 mM 2-(N-morpholino)ethanesulfonic acid, 3 mM MgCl2, 1.38 M KCl, and 20 mM ethylene glycol tetraacetic acid) and then fixed and permeabilized in 4% paraformaldehyde (Electron Microscopy Sciences, Hatfield, PA), 1.5% bovine serum albumin (Thermo Fisher Scientific), and 0.5% Triton X-100 in cytoskeleton buffer for 15 min at 37°C. Coverslips were then rinsed three times in phosphate-buffered saline (PBS) and incubated with Alexa Fluor 488–phalloidin (1:800; Invitrogen), mouse anti-paxillin (1:400; Millipore), and rabbit anti-fibronectin (1:400; Sigma-Aldrich) overnight at 4°C. The coverslips were then rinsed three times in PBS and incubated for 1 h with an Alexa Fluor 647 donkey anti-mouse (1:800; Invitrogen) or Alexa Fluor 568 goat anti-rabbit (1:400; Invitrogen) secondary antibody. Coverslips were mounted on glass slides using the SlowFade Antifade kit (Invitrogen).
Micropatterning via deep-ultraviolet (UV) illumination on polyacrylamide gels was done according to Oakes et al. (2014). Briefly, a chrome-plated quartz photomask (Microtronics, Newtown, PA) was cleaned with water and wiped with 0.5 ml of hexane (Sigma-Aldrich). A polyacrylamide gel mixture (7.5/0.3% weight percentage of acrylamide/bis-acrylamide) was polymerized for 30 min between the photomask and an activated glass coverslip, yielding a gel with a shear elastic modulus of 8.6 kPa. Once the gel was polymerized, the photomask was placed in a UVO-Cleaner 342 (Jelight, Irvine, CA) and illuminated with a combination of 185- and 254-nm UV light for 120 s. The coverslip and gel were then detached from the photomask by submerging the entire complex in water and gently lifting a corner of the coverslip with a tweezers. Gels were incubated in a solution containing 5 mg/ml EDC (Thermo Fisher Scientific) and 10 mg/ml NHS (Thermo Fisher Scientific) for 15 min. The EDC-NHS solution was then aspirated and replaced with a solution containing 10 µg/ml fibronectin in a buffer of 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES; pH 8.5) for 20 min. Gels were washed with PBS and incubated overnight in DMEM before cells were plated.
Microscopy and live-cell imaging
For time-lapse imaging of migration, coverslips were mounted in a custom-made chamber made from galvanized aluminum with six 22-mm-diamater holes. Vacuum grease was used to affix several round coverslips to this chamber to facilitate time-lapse imaging of multiple coverslips simultaneously. Imaging media was supplemented with 10 mM HEPES and 30 ml/ml Oxyrase (Oxyrase, Mansfield, OH). Imaging of migration and spreading was performed on a temperature-controlled inverted Nikon (Melville, NY) Ti-E microscope with a Lumen 200 Pro light source (Prior) and an HQ2 cooled charge-coupled device (CCD) camera (Roper Scientific, Trenton, NJ) controlled via MetaMorph acquisition software (Molecular Devices, Eugene, OR). Images were acquired using a 20 or 10×/0.75 numerical aperture (NA) Plan Fluor air objective (Nikon).
Fluorescence images (fixed and live-cell imaging) were obtained on an inverted microscope (Ti-E; Nikon) with a confocal scan head (CSU-X; Yokogawa Electric, Musashino, Tokyo, Japan), laser merge module containing 491-, 561-, and 642-nm laser lines (Spectral Applied Research, Richmond Hill, Canada), and HQ2 CCD camera (Roper Scientific). MetaMorph acquisition software was used to control the microscope hardware. Images were acquired using a 60×/1.2 NA Plan Apo water-immersion objective, a 20×/0.75 NA Plan Fluor multi-immersion objective (Nikon), or a 20× Plan Fluor air objective.
Analysis of cell shape and motion
F-actin images were thresholded in MATLAB to obtain a mask representing the cell. The built-in “regionprops” function was used to fit an ellipse to each cell mask and obtain its major (b) and minor (a) axes. Cell shape elongation was defined by the ratio b/a. Orientation was defined as P = cos2θ, where θ is the angle between a line parallel to the ECM and the long axis, b, of the cell ellipse. The average was calculated over all of the cells on a given ECM pattern imaged for each condition (N > 50).
Cells were manually tracked using the manual tracking plug-in in ImageJ. Instantaneous speed was measured from cell displacements at 20-min intervals. Mean cell speed was calculated by averaging instantaneous cell speed over a period of 10 h. To assess directionality of migration, we analyzed cell trajectories over a period of 10 h. For each time step in a trajectory, we calculated the angle θ between the displacement vector of a cell and the ECM. We defined a guidance parameter g such that for every displacement in a cell’s trajectory. Once we obtained guidance values, we calculated the mean guidance as a function of lag time:
We reported the guidance value obtained at τ = 300 because around this value the curves ceased to increase (Supplemental Figure S1B).
To determine the area and direction of protrusive fronts, we obtained contours of each cell by thresholding GFP-stargazin imaging. From the contours, we extracted protrusions using custom MATLAB software to identify the regions of area growth from between contours at times t and t + 2 min. Once protrusive regions were identified, we determined their orientation with respect to the ECM pattern by calculating P = cos2θ, where θ is the angle between a line parallel to the ECM and the line defined by the centroid of the cell and the centroid of the protrusive region. The average was taken over the angles of all cellular protrusions that occur in a 30-min time interval. Because ECM patterns are symmetric with respect to the y-axis, angles > 90° were reflected across the y-axis such that θ ∈ [0, 90]. Areas of protrusions were obtained using the regionprops function in MATLAB.
Statistical analysis and data reproducibility
The data for all the plots shown, with the exception of Supplemental Figure S3, D, E, G, and H, were obtained from at least two independent experiments. Supplemental Table S1 shows the number of replicates for the data in each figure. There were no limitations on data reproducibility. Exact sample sizes (N values) for all of the plots included in the figures are shown in Supplemental Table S2. Pairwise t tests were used to determine whether variables measured at different line spacings were statistically significant from each other. Post hoc analysis using the Holm method was used to correct for multiple pairwise comparisons. Welch’s t tests were used to determine whether the data from two experimental conditions at the same line spacing were significantly different. Exact p values for all tests are shown in Supplemental Table S3. All statistical analyses were performed using R software (RStudio Version 1.0.136).
This article was published online ahead of print in MBoC in Press (http://www.molbiolcell.org/cgi/doi/10.1091/mbc.E16-11-0769) on February 22, 2017.
guanine nucleotide exchange factor
green fluorescent protein
We thank Yvonne Beckham for assistance with molecular biology. We are also grateful to Barbara Hissa and Bob Harmon for assistance with Western blotting and to Tim Fessenden for critical comments on the figures. This work was supported by a National Science Foundation Graduate Research Fellowship to G.R.R-SJ. and National Institutes of Health Grant R01GM104032 to M.L.G.