The interplay of cell–cell and cell–substrate adhesion in collective cell migration

Abstract

Collective cell migration often involves notable cell–cell and cell–substrate adhesions and highly coordinated motion of touching cells. We focus on the interplay between cell–substrate adhesion and cell–cell adhesion. We show that the loss of cell-surface contact does not significantly alter the dynamic pattern of protrusions and retractions of fast migrating amoeboid cells (Dictyostelium discoideum), but significantly changes their ability to adhere to other cells. Analysis of the dynamics of cell shapes reveals that cells that are adherent to a surface may coordinate their motion with neighbouring cells through protrusion waves that travel across cell–cell contacts. However, while shape waves exist if cells are detached from surfaces, they do not couple cell to cell. In addition, our investigation of actin polymerization indicates that loss of cell-surface adhesion changes actin polymerization at cell–cell contacts. To further investigate cell–cell/cell–substrate interactions, we used optical micromanipulation to form cell–substrate contact at controlled locations. We find that both cell-shape dynamics and cytoskeletal activity respond rapidly to the formation of cell–substrate contact.

1. Introduction

Cells collectively migrate to carry out key functions of life: from tissue formation and organ development, to immune response and wound healing. Collective migration may simply involve a large number of cells flocking towards a common attractant, e.g. as observed during immune response. Collective migration may additionally involve cohesive migration patterns, in which cells migrate in a coordinated manner. Examples of such collective migration include human mammary epithelial cells, which form cohesive migration sheets during wound healing [1], and posterior lateral line primordial cells, which migrate in cohesive clumps during the development of the mechanosensory lateral line organ in Zebrafish [2]. Both cancer cells and amoeboid cells often migrate as multicellular streams, in which cells move in a head-to-tail fashion using the same path within tissues [3–5].

Collective migration requires the coordination of cells during migration, and therefore the coordination of the cell cytoskeleton and chemical signalling (e.g. chemotactic signalling) that control that migration. The chemical signalling cues that guide collective migration have extensively been studied, and many pathways have been identified. For example, human breast cancer cells have been found to form invading streams in vivo that use paracrine and autocrine signalling loops [6,7]. Dictyostelium discoideum cells provide a valuable model system for collective migration, since they form head-to-tail multicellular streams during aggregation and are guided by the chemoattractant cyclic adenosine monophosphate (cAMP) [4,8].

In addition to chemical guidance of cell migration, cells are also guided by mechanical cues. Mechanical forces arise from the extracellular matrix (ECM) through cell-surface contact and from other cells through cell–cell contact. These forces regulate the motion of migrating cell groups [9,10]. Cells are able to follow gradients in stiffness of the extracellular matrix (ECM), a phenomenon known as durotaxis [11]. In addition, cells can be guided by external physical forces exerted more locally by other cells or objects [10,12,13]. For example, it has been shown in vivo that fibre-like structures in the ECM can provide directional guidance and direct multicellular streams [3,9]. We previously showed that cell-surface adhesion can also affect collective migration: D. discoideum cells exhibit different collective migration patterns on surfaces with different inherent adhesivities [14]. However, it is not well understood how cell-surface adhesion affects collective migration, or how touching cells achieve highly coordinated motion.

This study focuses on the interplay between cell–cell and cell–substrate contact in migrating cells. Recent studies have shown that in epithelial cells these two adhesion systems spatially inhibit each other and use different mechanisms to regulate the cytoskeleton and to generate mechanical forces [15]. Epithelial cells and many other mammalian cells adhere to each other and to the substrate via integrins, the activation of which triggers signalling pathways that affect various cell behaviour [16]. On the other hand, some fast migrating cells, such as D. discoideum, adhere to surfaces non-specifically, and therefore allow us to study the competition of adhesions in the context of non-integrin-mediated adhesion.

We begin our investigation of the competition between cell–cell contact and cell-surface contact by blocking the formation of cell-surface contact. Without cell-surface contact, cells significantly change their collective migration pattern. However, the protrusions and retractions of individual cells are not significantly altered by inhibiting cell-surface contact, which is consistent with our previous observations of cells suspended in water [17]. Next, we investigate how cell-surface coupling affects cell–cell coupling and the spatial distribution of actin polymerization. Further, we use optical tweezers to form controlled cell–substrate contacts, which allows us to systematically study how cell-shape dynamics and the cytoskeleton respond to changes in cell–cell and cell–substrate contact.

2. Results and discussion

2.1. Without cell-surface contact, Dictyostelium discoideum do not stream in a head-to-tall fashion but instead aggregate by clumping

We used two complementary approaches of inhibiting cell-surface contact in order to evaluate the effects of cell–substrate adhesion on cell–cell adhesion. In our first approach, wild-type cells (AX3) were plated and remained suspended on a polyethylene-glycol (PEG)-coated surface (MicroSurface Inc., MO, USA). PEG coatings have previously been used to prevent cells from adhering to surface [18]. Interference reflection microscopy (IRM) [19] was used to determine the actual cell-surface contact area. Bright-field and IRM images of AX3 cells on glass are shown in figure 1a. As we previously reported [14], cells partially adhere to glass surfaces. (Regions of adherence appear dark in IRM images. See figure 1a for an example.) On PEG-coated surfaces, cells are less polarized and do not form regions of cell-surface adhesions as shown in figure 1b (no dark region in the IRM image).

Figure 1.

On PEG-coated surfaces, cells exhibit no cell-surface contact and aggregate by clumping rather than streaming. (a) Bright-field and IRM images of AX3 cells show that cells plated on glass form cell-surface contacts. (Dark areas on the IRM image indicate regions of cell-surface contact.) (b) Bright-field and IRM images of AX3 cells plated on a PEG-coated surface show that cells do not form cell-surface contacts on PEG. (No dark areas are observed on the IRM image.) (c) Early aggregation of AX3 (WT) cells on glass; cells aggregate via multicellular streaming. (d) Early aggregation of AX3 (WT) cells on a PEG-coated surface; cells aggregate by clumping instead of streaming. (e) Extracted motion tracks of active cell movement (corrected movement) on PEG-coated surfaces. (f) A multicellular stream formed by AX3-Lifeact-RFP cells with cells aligned head-to-tail on glass. (g) A multicellular clump formed by AX3-Lifeact-RFP cells on a PEG-coated surface.

With these two surfaces, we examine the migration of D. discoideum cells. We investigate cells at an early aggregation stage, where cells are prone to signal and to each other and migrate collectively in a head-to-tail fashion. Cells were marked with the cytosolic stain CellTracker Green (Invitrogen) to facilitate the imaging and analysis of dynamic changes in cell shape. Representative images and movie are shown in figure 1c,d and electronic supplementary material, movie 1. On glass, cells are initially uniformly distributed on the surface and move non-directionally. After the first 20 min, the cAMP secreted by cells facilitates the formation of multicellular streams. This process is well established as a key example of collective streaming [4]. Collective streaming results in the formation of a few large cell aggregates.

By contrast, cells plated on PEG-coated surfaces do not stream collectively. Instead, they move non-directionally and form small spherical aggregates (figure 1d and electronic supplementary material, movie 1). After several hours, these spherical aggregates merge into larger aggregates. Since cells remain suspended on PEG-coated surfaces, their movement is largely affected by the convection and flows in the chamber. Therefore, cell movement is actually the combination of passive movement that caused by environment factors and active movement that results from their aggregation motion.

To distinguish between active and passive movements, we used a template matching plugin in ImageJ software (National Institutes of Health; http://rsbweb.nih.gov/ij/) to get rid of the passive movement of all cells. Then, a custom particle tracking Matlab (The Mathworks, Natick, MA, USA) code was applied to obtain the movement of each cell or cell clump, from which we calculated the active movement of cells in the field of view. Electronic supplementary material, movie 2 and figure S1 show the comparison of the extracted motion tracks between overall cell movement (original movement) and the active cell movement (corrected movement). After subtracting the passive movement, corrected movement tracks clearly show the aggregation of cells towards an aggregation centre (figure 1e).

We found although the overall speed of cells with/without passive motion are similar (14.9 μm min⁻¹ for original motion; 14.6 μm min⁻¹ for corrected motion), subtracting the passive flow made the speed more uniform over time (electronic supplementary material, figure S2).

To observe actin polymerization dynamics, we use AX3-Lifeact-RFP cells (F-actin labelled with red fluorescent protein). Figure 1f,g is higher magnification images of AX3-Lifeact-RFP cells, which illustrate that cells arrange in streams on glass surfaces and in spherical aggregates on PEG-coated surfaces. During collective streaming on glass, adherent cells align in a head-to-tail fashion in streams. However, suspended cells do not exhibit such alignment inside aggregates. Furthermore, cell–cell contacts in cells that are adhered to a surface show enhancement of F-actin at cell–cell contacts, as shown in figure 1f. In cells that are not adhered to a surface, actin is instead enhanced in the periphery of the aggregate, as shown in figure 1g.

2.2. Inhibiting cell–cell contact results in significant collective streaming defect

We examine how cell–cell adhesion affects collective migration in the presence of cell-surface adhesion by using the force of electrostatic repulsion between cells. This approach has previously been used and adapted by us to prevent adhesion between cells and glass coverslips [20]. Since both cell membranes and glass are negatively charged, they repel one another. Under usual experimental conditions, these negative charges are screened by ions in the phosphate buffer (PB). Repulsion between cells increases as the ion concentration in the medium decreases. In pure water and medium with low ion concentration, this repulsion is strong enough to prevent or partially inhibit cell–cell and cell-surface adhesions. Although this experimental approach only works for our osmotic-shock-resistant amoeboid cells and is not transferrable to mammalian cells, it allows us to tune the mechanical cell–cell contact and elucidate the potential role of cell–cell adhesion in collective migration.

In these experiments, cells were allowed to form cell-surface adhesion in a diluted medium (6% PB) for 15 min. Then, the medium was adjusted to different concentration (from full medium (100% PB) to almost pure water (1% PB)) to tune cell–cell adhesion at a different level, as shown in the electronic supplementary material, movie 3. We found that diluted PB (3%–10%) greatly inhibits the formation of cell–cell contact but retains cell-surface contact and cell motility. Cells in the 100% PB form streams normally (figure 2a); cells exhibit significant streaming defect and aggregate through very short streams when cell–cell contact is partially inhibited (6% PB) (figure 2b); cells in the medium with even lower ion concentration (3% PB) remain active motion but do not form multicellular streams due to the significant inhibition of cell–cell contact (figure 2c); when we further dilute the medium to 1% PB, cell-surface adhesion is also inhibited and some cells detached from the surface (figure 2d). Testing via internal reflection microscopy (IRM) revealed that cells form cell-surface contacts in the first three conditions only (data not shown). Therefore, these experiments demonstrate the importance of cell–cell adhesion for collective streaming migration: cells do not align in a head-to-tail fashion without cell–cell adhesion.

Figure 2.

Inhibiting cell–cell contact result in significant collective streaming defect. (a) Migration of AX3 (WT) cells on glass in 100% PB; cells aggregate via multicellular streaming. (b) Migration of AX3 (WT) cells on glass in 6% PB; cells aggregate with streaming defect due to the partially inhibition of cell–cell contact. (c) Migration of AX3 (WT) cells on glass in 3% PB; cells do not form streams or aggregate. (d) Migration of AX3 (WT) cells on glass in 1% PB; some cells lost cell-surface adhesion. (e) AX3 cells without cell–cell and cell–substrate adhesion exhibit dynamic and polarized morphologies.

We further study the extreme case in which both cell–cell and cell-surface adhesion are inhibited. Cells were placed in distilled (DI) water instead of ionized buffer (PB), in which condition the electrostatic repulsion is strong enough to prevent any cell–cell and cell-surface adhesion. Cells suspended in DI water lose both cell-surface adhesion and cell–cell adhesion. However, as shown in figure 2e and electronic supplementary material, movie 4, these cells retain their polarity and motility (as well as their cytoskeletal activity, data are not shown).

Cells that are suspended on PEG-coated surfaces have cell–cell adhesion but no cell-surface adhesion. Although they retain their shape activities, they cannot form streams since they lose the head-to-tail alignment. On the other hand, cells that are adhered on surfaces display significant multicellular streaming defect when the cell–cell adhesion is inhibited. Therefore, collective streaming requires both cell–cell adhesion and cell-surface adhesion.

2.3. The loss of cell-surface contact does not significantly change the shape dynamics and actin activity of individual cells

Both cell–cell adhesion and cell-surface adhesion are important in collective migration. To understand their respective roles, we first observe cell-surface adhesion in the absence of cell–cell adhesion, i.e. how cell-surface adhesion impacts on individual cells. Since cell migration involves a careful interplay of protrusions and retractions, quantitative analysis is used to compare the motion of individual, adherent cells with the motion of individual, suspended cells. Our previously developed approach represents the shapes and shape dynamics of cells in a way that provides a whole-cell perspective on the protrusive and retractive processes [17,21]. Briefly, the outline of an individual adherent cell is extracted from a sequence of images. To visualize local protrusions, the curvature at each boundary point is calculated and represented as a different colour (figure 3a). Red represents convex regions, whereas blue represents concave regions, i.e. invaginations. Electronic supplementary material, movie 5b shows an individual cell migrating on a glass surface, with its boundary, coloured by curvature, overlaid. The shape dynamics of this cell can then be represented as a kymograph of the local boundary curvature. A representative case is shown in figure 3b. The shape at each time is represented by a vertical line of equal length that is coloured by local curvature. The two horizontal red regions indicate the locations of the front and back (the two polarized ends) of the cell and the slanted red lines indicate protrusion waves travelling on the sides of the cell. We define the polarized end that protrusion waves initiate from as the ‘front’ of the cell and the other end as the ‘back’. Figure 3c,d and electronic supplementary material, movie 6b are a representation of the curvature outline and the cell-shape dynamics of an individual cell that is suspended on a PEG-coated surface. From figure 3b,d, we can infer that both the adherent cell and the suspended cell are polarized, i.e. have a well-defined front and a back, with protrusion waves that travel from the front to the back of the cell. Protrusion waves, which are indicated by black dashed lines, are seen as slanted red lines that initiate at the cell front and propagate to the back. Polarized shapes and similar pattern of protrusion waves are observed on both adherent cells and suspended cells. Further analysis of protrusion waves finds that curvature waves on suspended cells are more frequent than those on adherent cells (see the electronic supplementary material, figure S3).

Figure 3.

Cells without surface contact exhibit similar shape dynamics and F-actin activity as cells with surface contact. (a) An image of a representative adherent AX3-Lifeact-RFP cell with the cell boundary, coloured by curvature, overlaid. Here, red colour represents regions of positive curvature, whereas blue colour represents regions of negative curvature. (b) A kymograph of boundary-point-curvature versus time for an individual, adherent cell. (c) An image of a representative suspended AX3-Lifeact-RFP cell with the cell boundary, coloured by curvature, overlaid. (d) A kymograph of boundary-point-curvature versus time for an individual, suspended cell. These kymographs illustrate that adherent and suspended cells have similar cell-shape dynamics. Both of them have two horizontal red region (front and back of the cell) and slanted red lines start from the front red region and end at the back red region (indicated by black dashed lines), which correspond to curvature waves travel on cell boundary from cell front to the back. (e) An image of a representative adherent AX3-Lifeact-RFP cell with the cell boundary, coloured by the intensity of nearby actin fluorescence, overlaid. Red represents region of high fluorescence intensity, and blue represents regions of low fluorescence intensity. (f) A kymograph of boundary-point-fluorescence-intensity versus time for an individual adherent cell. (g) An image of a representative suspended AX3-Lifeact-RFP cell with the cell boundary, coloured by the intensity of nearby actin fluorescence, overlaid. (h) A kymograph of boundary-point-fluorescence-intensity versus time for an individual suspended cell. These kymographs illustrate that adherent and suspended cells show similar patterns of F-actin activity. Regions of high actin polymerization (shown red) mainly occur at the front of cells but sometimes travel in a wave-like fashion along the sides of cells (indicated by black dashed lines).

To investigate the intracellular processes that allow cells to maintain their polarity and motility, we visualize intracellular actin polymerization using Lifeact-RFP. As shown in the electronic supplementary material, movies 5a and 6a, actin polymerization is very dynamic at the leading edge of migrating cells. Polymerization bursts are very transient. F-actin dynamics is measured by analysing the fluorescence intensity of Lifeact-RFP along the cell periphery. In figure 3e,g, electronic supplementary material, movies 5c and 6c, the colour on the extracted outline represents the F-actin fluorescence intensity near each boundary point. Red represents high fluorescence intensity (high F-actin concentration), whereas blue represents low fluorescence intensity (low F-actin concentration). The kymographs in figure 3f,h show the F-actin activity of representative individual cells that are migrating on a glass surface and are suspended on a PEG-coated surface, respectively. The red areas represent regions of high actin polymerization, which mainly occur at the front of the cell and travel in a wave-like fashion mostly along the sides of the cell that near the front (indicated by black dashed lines). Figure 3f,h suggests that both adherent and suspended cells dynamically and asymmetrically assemble actin to generate force on their boundary, so as to maintain their shape polarity and their leading edge, where the membrane is pushed outwards.

2.4. The shape dynamics of pairs of cells is coupled in a manner that depends on cell-surface contact

Without cell-surface contact, D. discoideum cells maintain their cell-shape dynamics and cytoskeletal activity when they move as individuals. But, as shown in figure 1d, large groups of cells cannot form normal collective migration patterns without cell-surface adhesion. We hypothesize that the loss of cell-surface contact changes the mechanical interaction between cells, which results in the differences in collective behaviour. Thus, pairs of cells in direct contact were studied, as they represent the simplest case of cell–cell interaction.

When two neighbouring cells move in a stream on glass, their cell boundaries are sometimes indistinguishable. To distinguish neighbouring cells, a mixture of AX3 cells stained with either CellTracker Green (Invitrogen) or CellTracker Orange (Invitrogen) was used. Figure 4a and electronic supplementary material, movie 7 show a neighbouring pair of differently stained cells migrating on glass. For this analysis, we term the cell that is ahead as the ‘leading cell’, and the cell that is behind as the ‘trailing cell’. The overall boundaries of pairs of cells were extracted. Figure 4b shows overlaid cell shapes from different points in time (6 s between shapes). It indicates protrusions travel in the same direction on both cells and cross cell–cell contact region. Protrusions near the cell–cell contact region are pointed out by coloured arrows. Further, the shape dynamics of each pair was measured following the approach used for individual cells (figure 4c). Note that the two red horizontal lines represent the front of the leading cell (cell A) and the back of the trailing cell (cell B), and the slanted red lines indicate protrusions travelling on the side of these cells. As observed for individual cells, travelling protrusions initiate at the front of the leading cell. For cell pairs, these protrusion waves travel across the contact region of the two cells (indicated by the purple arrow), and stop when they hit the back of the trailing cell. This suggests that membrane protrusion waves on both cells travel in the same direction and are synchronized in time, i.e. the shape dynamics of these two cells are coupled across the cell–cell contact. For further comparison, we sketch a simplified cartoon of the shape dynamics in figure 4d. Overlapping extracted cell outlines indicate the generation of a membrane protrusion, and the arrows indicate the direction of protrusion propagation.

Figure 4.

Adherent cells exhibit coupled protrusion waves but suspended cells do not. (a) An image of a representative pair of AX3 cells that are head-to-tail aligned and migrating on glass. (Cells were cytosolically stained green and orange). (b) Overlay of the shapes of this pair of cells, which shows that the protrusion waves in both cells travel in the same direction and cross the cell–cell contact region. Red outline is a cell shape extracted from an earlier frame, whereas blue outline is a cell shape extracted from a later frame. Protrusions are pointed out by arrows. (c) A kymograph of boundary-point-curvature versus time for this pair of adherent cells. Cell boundaries were extracted from the overall boundary of the two cells. The kymograph shows that curvature waves, indicated by black dashed lines, travel along the same direction in these two cells. (d) A picture of the shape dynamics of a pair of adherent cells. Protrusion waves in different cells propagate in the same direction and travel from cell to cell. (e) An image of a representative pair of AX3 cells suspended over a PEG-coated surface. (Cells were cytosolically stained.) (f) Overlay of the shapes of this pair of suspended cells, which shows that protrusion waves travel in opposite directions along the two cells. (Protrusions are pointed out by arrows on both cells, from red to blue). (g) A kymograph of boundary-point-curvature versus time for this pair of suspended cells. Cell boundaries are extracted from the overall boundary of the two cells. The kymograph shows that the curvature waves (indicated by black dashed lines) travel in opposite directions along these two cells. (h) A picture of the shape dynamics of a pair of suspended cells, protrusion waves in different cells propagate in opposite directions and do not travel from cell to cell. (i) Cell–substrate contact can induce a switch in cell polarity. Upon contact with a bead at its front, this cell rounds up, and then forms a new front away from the bead.

Pairs of suspended cells exhibit different collective motion and shape dynamics than pairs of adherent cells (figure 4e,f,g and electronic supplementary material, movie 8). Overlaid cell shapes (figure 4f) indicate that protrusions travel in opposite directions towards cell–cell contact region on pairs of cells. Coupled, suspended cells do not form cell–cell contact in a head-to-tail fashion. Instead, they often form tail-to-tail contact, as shown in figure 4g. In this boundary curvature kymograph of a representative pair of cells, the two horizontal red lines represent the front of the two cells and the slanted red lines represent travelling protrusion waves. The cells touch back-to-back, and protrusion waves initiate at the fronts of both cells, travel laterally along the cells, and then stop at the region of cell–cell contact. A schematic of the dynamics of protrusions in suspended cells is shown in figure 4h. In general, pairs of suspended cells generate new protrusions at the end that is farthest away from the region of cell–cell contact. Those protrusions propagate in opposite directions and do not travel across the cell–cell contact. The comparison between pairs of adherent cells and suspended cells suggests that, without cell-surface contact, cells cannot form the head-to-tail cell–cell contact that helps them to synchronize their motion.

To further investigate the response of suspended cells to contact with another object, we use controlled indirect optical-gripping of cells, a technique that we had previously developed and refined [22–25]. In our sample applications, we demonstrated that assistant tools, such as glass beads, could be used to indirectly bring two cells in controlled contact, without directly exposing cells to the potential photo-damage from the laser. Surprisingly, we had found that two suspended cells that are brought into front-to-back contact do not retain that alignment but instead rapidly change the location of their leading edges [22]. One possible reason is that the loss of cell-surface adhesion varies the adhesivity of the cell membrane, e.g. the leading edge or tail of cells becomes less sticky and so cells are more likely to adhere at other positions. To analyse the role of cell–substrate contact in a controlled way, we move glass beads (5 μm in diameter, Bangs Laboratories, Fishers, IN, USA) towards target cells using optical tweezers (figure 4i and electronic supplementary material, movie 9). Once the cell–bead distance is small enough to form direct cell–bead contact, we release the bead from the laser trap. Therefore, there is no optical tweezer force that pushes or pulls the cell after the formation of cell–bead contact. When a glass bead is adhered to the leading edge of a suspended cell, the cell rapidly loses its polarity and generates a new leading edge far from the cell–bead contact region, consistent with our observations of suspended cell behaviour in response to cell–cell contacts. Our observations suggest that cells actively alter their shape dynamics and leading edge locations when forming direct contact with other objects, such as beads or cells.

2.5. The spatial patterning of actin polymerization in directly contacted cells varies with cell–substrate contact

We investigate how actin polymerization responds to changes in cell–cell contact, and how this response differs for adherent and non-adherent cells. The F-actin concentration of a representative pair of cells migrating with front-to-back alignment on glass is shown in figure 5a and electronic supplementary material, movie 10. Actin mainly polymerizes at the front of the leading cell and at the contact region of the two cells. The fluorescence in the contact region is mostly due to the enrichment of F-actin at the leading edge of the trailing cell, as shown in the electronic supplementary material, figure S4. The kymograph of the fluorescence intensity near the boundary of this pair of cells indicates that this enrichment near the leading edge and at cell–cell contacts persists during migration (figure 5b). Statistical analysis of the relative fluorescence intensity on the cell boundaries (figure 5c) supports the observation that the cell–cell contact region exhibits significantly higher F-actin activity than the far edges (free ends) of either cell. (The coloured rectangular boxes in figure 5a indicate the regions that are analysed in figure 5c. The rectangular boxes in figure 5d,f,g,i,j also indicate regions of analysis).

Figure 5.

The spatial distribution of F-actin in pairs of cells responds to a new cell-surface contact. (a) A fluorescence image of two AX3-Lifeact-RFP cells in direct contact on glass. Coloured, rectangular boxes indicate the regions that are analysed in (c). (b) A kymograph of boundary-point-fluorescence-intensity versus time for this pair of adherent cells. F-actin is persistently enriched at cell–cell contacts. (c) Relative F-actin fluorescence intensity in different regions on the boundary for the pair of adherent cells. The cell–cell contact region (shown red) shows higher fluorescence intensity than the free ends of the cells (shown in blue and yellow colours). (d) A fluorescence image of two AX3-Lifeact-RFP cells in direct contact suspended on a PEG-coated surface. Coloured rectangular boxes indicate the regions that are analysed in (f). (e) A kymograph of boundary-point-fluorescence-intensity versus time for this pair of suspended cells. No enrichment of F-actin is found at the cell–cell contact. (f) Relative F-actin fluorescence intensity in different regions on the boundary for the pair of suspended cells. The cell–cell contact region (shown in red colour) shows lower fluorescence intensity than the free ends of the cells (shown in blue and yellow colours). (g) Bright-field and fluorescence images of a suspended AX3-Lifeact-RFP cell with a 5 μm silica bead nearby. (h) Using optical tweezers, the silica bead is adhered to a recently generated protrusion on the cell, and sticks on the membrane of the cell. (i) Bright-field and fluorescence images of the cell after manipulation show that, following contact with the bead, F-actin is enriched at the end of the cell opposite to the cell–bead contact region. (j) Relative F-actin fluorescence intensity in different regions on the boundary of a cell before and after the formation of cell–bead contact. (In c,f,j, to make intensity of each frame comparable, the average intensity of the whole-cell boundary in each frame was normalized to be 1. Each analysed region includes the same number of boundary points. Error bar: standard deviation. *: T-test p < 0.001).

Similarly, the F-actin concentration in pairs of suspended cells that display back-to-back alignment was investigated (figure 5d and electronic supplementary material, movie 11). In these cells, actin mostly polymerizes at the far edges (free ends) of both cells, with no significant F-actin enrichment in the cell–cell contact region. Kymographs of F-actin intensity (figure 5e) and statistical analysis of relative fluorescence intensity (figure 5f) further validate this observation.

Our results suggest that, without cell-surface adhesion, cells organize their cytoskeleton differently when they adhere to other cells. From analysis of shape dynamics and F-actin activity, we find that in both adherent and suspended cells, actin polymerizes at the cell fronts, where protrusion waves are initiated. However, in pairs of suspended and adherent cells, the cell fronts occur in different locations. Adherent cells locate their fronts at the tails of other cells, whereas suspended cells locate their fronts away from other cells.

Using optical micromanipulation as described earlier in the paper, we next investigate how the introduction of controlled cell-surface contacts drives reorganization of the intracellular cytoskeleton. To simultaneously optically manipulate beads and measure actin fluorescence, we built a florescence light path into our optical tweezers system. This allowed us to image the spatial distribution of F-actin within cells before and after beads were adhered onto cells via optical micromanipulation. The typical process of such manipulation is shown in figure 5g–i and electronic supplementary material, movie 12. We find that upon contact with a silica bead, F-actin is enriched away from the region of cell–bead contact. This is consistent with and provides a molecular explanation for our observation that protrusions occur away from regions of cell–bead contact (figure 4i). Further analysis of the F-actin intensity along a cell boundary is shown in figure 5j. This analysis further validates that cells form a leading edge away from cell-surface contact and that F-actin is enriched at that leading edge. Intracellular actin polymerization actively responds to the adhesion of either a silica bead or another cell in a similar manner.

3. Conclusion

Our main result is that the presence of cell-surface contacts affects actin polymerization at cell–cell contacts, and that cell-surface contacts are important to facilitate head-to-tail collective migration. Even though the shape dynamics and internal actin polymerization dynamics of individual cells during amoeboid migration do not appear to depend on surface adhesion, we find that the loss of surface adhesion significantly alters the collective migration pattern. Our analysis suggests that this change in collective behaviour is not simply due to the fact that cells are not able to move well without surface adhesion (though they are able to ‘swim’ as reported by us and others in prior work [17,26,27]), but due to changes in the shape dynamics.

The motor underlying this shape dynamics is studied through observing actin polymerization activity. Our results indicate that individual suspended cells display similar F-actin activity as adherent cells, but that actin polymerization in touching cells is only coordinated to facilitate motion in the same direction if a cell-surface contact is present. The loss of cell-surface contacts leads to a loss of the cell–cell mechanical coupling via protrusion waves and actin polymerization enrichment. To generate a controlled cell–substrate contact, and to image intracellular actin polymerization at the same time we adapted our optical micromanipulation system. We find that after forming controlled cell–substrate contacts cells respond to the additional contact points with enhanced actin polymerization.

These observations from the optical manipulation experiments give us complementary insights into mechanotransduction in two aspects: (i) as previously shown, motility or shape change can be stimulated by forming contact with objects that are coated with certain chemical components (e.g. cadherin, antibody or other protein/protein complexes) [12,28,29]. The mechano-chemical effects reported in these studies are specifically mediated by the chemical components that are coated on the objects. Whereas, since we used uncoated silica beads and D. discoideum cells adhere to substrate non-specifically, i.e. no integrin-mediated adhesions, the changes in cell shapes dynamics and the reaction of actin activity we report in our study are not chemical component-specific. (ii) Forces in the range of hundreds of piconewtons to hundreds of nanonewtons exerting on cellular structures appear needed to activate mechano-chemical signalling pathways, known as mechanotransduction pathways [12,29–31]. However, in our experiments, we observed changes of cell actin dynamics without the need to apply any external forces on cells, i.e. forming cell-surface contact can induce a change in intracellular dynamics. It would be interesting to test whether applying forces can induce different intracellular responds on this system. Such experiments require the use of different force spectroscopy that can exert higher forces than optical tweezer (e.g. magnetic tweezer or AFM), which can exert up to 10 nN forces [30,32].

Collective aggregation is an important behaviour of cell groups. Studies of collective migration generally focus on chemical signalling between cells, and the migration of cells in response to that signal. Our results elucidate that, in addition, cell-surface coupling can facilitate collective migration via coupling of actin polymerization activity. On the other hand, when cells are not contact with a surface, they are no longer able to migrate in a cooperative manner. Indeed, we observe less polymerization activity at cell–cell contacts but instead protrusions and actin polymerization at the boundaries of the aggregate. Though this is a very distinct localization of mechanical activity, it also leads to aggregation of cells into larger groups, very similar to collective migration on an adhesive surface. Our findings suggest that aggregating cells do not simply utilize actin polymerization to develop protrusions in response to chemotactic signals. Our data suggest that actin polymerization dynamics is also coordinated by cell­–cell adhesion and cell-surface adhesion in a way that facilitates collective behaviour. The migratory machinery of the cell thus appears to weigh and balance mechanical stimuli with chemotactic signals in yet unexplained ways to facilitate robust aggregation.

4. Material and methods

4.1. Model system

To study the interplay between mechanical and biochemical signalling, we use a simple model system, the social amoebae D. discoideum, which form multicellular streams during development [4]. They do so by simultaneously chemotaxing towards and secreting adenosine 3′,5′-cyclic monophosphate (cAMP), thereby forming aggregates which eventually differentiate into fruiting bodies. When exposed to a gradient of cAMP, the cells quickly orient themselves up the gradient and migrate using F-actin-based pseudopodial extensions of their front, coupled with myosin II-mediated contraction and retraction of their sides and back. Cyclic AMP not only induces cells to migrate directionally, but also stimulates the cells to produce and release cAMP locally, which allows cells to relay the chemoattractant signal to distal cells and migrate collectively in a characteristic head-to-tail fashion.

4.2. Tissue culture, differentiation and labelling

Wild-type D. discoideum (AX3) cells and the mutant AX3-Lifeact-RFP cells were grown in HL-5 medium at 21°C to concentrations not higher than 5 × 10⁶ cells ml⁻¹ [33]. For experiments, cells were starved and developed for 5 h in development buffer (DB: 5 mM KH₂PO₄; 5 mM Na₂HPO₄*7H₂O; 2 mM MgSO₄; 0.2 mM CaCl₂) and pulsed with 75 nM cAMP every 6 min, as described in other papers [34,35]. Developed cells were harvested after 5 h by centrifugation of 500 μl of the development liquid at 9000 r.p.m. for 3 min. The cell pellets were washed twice and dissolved in 500 μl of PB (5 mM KH₂PO₄; 5 mM Na₂HPO₄*7H₂O) or distilled water. In some experiments, cells were fluorescently stained with CellTracker Green CDMFA or CellTracker Orange CMTMR (Invitrogen) at concentrations of 18 μg ml⁻¹ for 20 min, similarly to previously reported procedures [35].

4.3. Microscopy

The early aggregation of AX3-Lifeact-RFP cells was imaged for 2.5 h every 6 s using a Leica SP2 confocal microscope with a 10× objective. A total of 4 × 10⁵ cells were well mixed in 300 μl of PB and added in a 2-well Lab-tek chamber 15 min prior to imaging to allow cells to adhere to the surface. Other images of AX3-Lifeact-RFP Cells were taken every 2.5 or 2 s using either a Zeiss 510 confocal microscope with a 40× objective or a Leica SP2 confocal microscope with a 100× objective.

Optical manipulation images were taken using a Nikon inverted light microscope with a 60× objective, which is integrated with a Biorryx optical trapping system (Arryx Inc., Chicago, IL, USA). Optical traps were generated with a 532 nm laser (Nd:YAG 5W, Spectra-Physics, Newport Inc., Irvine, CA, USA). A fluorescence system was also incorporated into the same microscope to facilitate fluorescence imaging (Ex: 540–580 nm, Em: 620–700 nm). Images were collected by a highly sensitive CCD camera (PCO. Edge, Kelheim, Germany).

4.4. Image analysis

Confocal images were processed by a custom-shape analysis Matlab (The Mathworks, Natick, MA, USA) program as described in previous publications [17,36]. Shapes of individual cells or pairs of cells were extracted and 400 points on the boundary were obtained on each frame. All boundary points can be tracked by applying 1 : 1 mapping between points on each frame and the following frame. The mapping is based on the finding of the minimum sum of square distance between points. For each boundary point, the curvature was calculated by fitting a circle into this boundary point and two points that are 10 points away from it. For visualization, the value of the curvature is scaled in colour in the kymograph with a cutoff at a maximum curvature magnitude (0.25). The colour bars are normalized by the cutoff (−2.5–2.5). For the fluorescent intensity measurement, we draw a circle (10 pixels in diameter) around each boundary point and calculate the average actin intensity within that circle. Then, the value of the intensity of each points was normalized by the maximum intensity on the cell boundary and scaled in colour in the kymograph.

Funding statement

This work was supported by the Intramural Research Program of the Center for Cancer Research, NCI, NIH as well as by the NSF grant nos. CPS0931508 and PHY1205965.