Identification of Leader and Follower Cells
To study the requirements for persistent cell protrusion we chose to analyze the migratory response of PtK1 cells undergoing wound repair. Similar to other epithelial cell types [22
], select PtK1 cells, referred to as leader cells, developed persistent protrusions in response to a wound inflicted on a confluent cell monolayer (). Time lapse imaging showed that wound edge cells required 1–2 hours to reestablish adhesive contacts to the glass substrate, and cells were seen to crawl toward the site of injury soon thereafter. Leader cells were visible 2–3 hours after wounding, residing at the tip of newly formed cell mounds flanked on either side by follower cells. Leader cells maintained their ‘lead’ position over many hours or until the wound was fully closed. They exhibited an extended lamella and a broad leading edge that rarely retracted (; Supplemental Data, Video 1
). In contrast, follower cells were characterized by compacted lamellae and sporadic protrusions that randomly alternated with phases of edge retraction.
Identification of leaders and followers during wound healing of PtK1 cells
Previous motility studies involving PtK1 cells have focused on islands composed of 2 – 8 cells [19
]. Island cells for the most part did not exhibit directed motility (). Instead, cell morphologies were varied. In some instances they resembled follower cells in that their leading edge would randomly cycle between protrusion and retraction. Other island cells exhibited phases of persistent protrusion, although they were short-lived relative to leader cells (minutes instead of hours).
Analysis of Leading Edge Activity
To analyze the different protrusion behaviors of these cell types, time-lapse images were acquired at 10 second intervals, over an observation window of 7–10 minutes. Edge movements were tracked along the entire cell edge with submicron resolution via automated edge detection software [26
] (; Supplemental Data, Video 2
). Superimposed edge positions indicate that in leader cells most sectors move forward over 10 minutes. This results in substantial net advancement of the entire cell edge. In contrast, over the same period individual boundary sectors in follower and island cells advance and retract in quasi-random cycles. Thus, the average cell position remained largely unchanged.
To quantify these observations we computed the ‘mean protrusion rate’, i.e. the total net displacement of the cell edge divided by the duration of the observation period, with the average obtained from n = 4 to 8 cells (, left panel; see Exp. Procedures). Leaders had two- to three-fold higher protrusion rates than follower and island cells. The same cell averaging procedure was applied to determine the ‘protrusion persistence,’ i.e. the time ratio an individual edge sector spends in advancement versus in retraction (see Exp. Procedures). Leader cells exhibited approximately three-fold higher persistence than followers (Figure 1E, center panel; p-value < 0.05). They also showed slightly faster ‘instantaneous protrusion velocities,’ i.e. the average of all (positive) protrusion velocities of all reporter windows (p-value < 0.05 between leaders and followers). However, the instantaneous retraction velocities were indistinguishable between the three cell types (, right panel; p-value > 0.3). Together, these analyses suggest that leader cells protrude efficiently because they advance more coordinately in time and space, and, to lesser extent, because they advance faster.
Cross-Correlation Analysis of Leading Edge and F-actin Dynamics
The instantaneous protrusion velocity is determined by the difference in the rates of F-actin assembly and retrograde flow at the leading edge [27
]. The latter depends on how the propulsive forces at the plasma membrane are balanced by adhesion formation [28
] and may also depend on the mechanical coupling between the lamellipodium and lamella [19
], defining the degree of contraction force transmission to the leading edge [29
]. The protrusion persistence is largely determined by the time scales over which the assembly-to-flow ratio varies. Therefore, to understand the differences in protrusion efficiency between cells, it was crucial to determine from independent measurements the relationships between edge movement, rates of filament assembly, and rates of retrograde flow.
To achieve this, we constructed activity maps on the basis of reporter windows (; windows 2–4 µm in width, 1–2 µm in depth) aligned along the entire cell edge (see Exp. Procedures). Within each window we determined the edge protrusion/retraction rates and locally
compared it to the activity maps of F-actin flow and turnover (). The strength of the relationship was determined by the magnitude of the pairwise cross-correlation between activity maps, including time-shifts to account for differences in the relative timing of activities (see Exp. Procedures)[30
]. Correlation functions extracted from individual reporter windows were then averaged along the cell edge and subsequently over multiple cells. Underlying this procedure is the assumption that the protrusion and retraction states may be heterogeneous along the cell edge (and over time), but the relationships between edge movement, filament assembly and flow would be preserved.
Relationships between actin filament dynamics in lamellipodium and protrusion/retraction activity
Using this analysis, leader cells showed significant correlation between protrusion and assembly at −20 s. This behavior was consistent amongst leaders (; top row – left panel), resulting in a strong averaged correlation profile (; top row – right panel). The time shift of −20s indicates that maximal F-actin assembly lags behind the fastest forward edge motion by ~20s [28
]. In contrast, cross-correlation between protrusion and retrograde flow (; mid row) resulted in a negative correlation score, reaching a minimum at −40s. Therefore, 40s after the maximal protrusion activity, retrograde flow of F-actin transiently speeds up. Consistent with this time course, the cross-correlation between F-actin retrograde flow and F-actin turnover led to a negative peak at +20s (, bottom row), indicating that the transient increase in retrograde flow is preceded by increased assembly by ~20s. Thus, in a leader cell, edge advancement is initiated by increased assembly of actin filaments, which reaches a maximum ~20s after cell undergoes fast protrusion. As F-actin reaches the state of maximal assembly, filaments begin to slide rearward, possibly because propulsion forces exceed the force levels sustained by substrate adhesion [28
]. The fastest flow rates occur ~20s after maximal F-actin assembly. This event terminates a protrusion cycle, which lasts 100 – 120s (derived from the temporal autocorrelation of the protrusion activity map).
Follower and island cells displayed distinctly different timing relationships. Overall, the significance of the correlation functions was weaker than with leaders, owing to increased cell-to-cell variation (; left columns). This suggested that protrusion in these cells is more complex, possibly the result of convoluted contributions from multiple mechanisms. Indeed, both cell types exhibited a transient increase in filament assembly during protrusion, albeit without the time delay found in leaders (; top row). Furthermore, they showed a significant positive correlation score at zero time lag between protrusion and retrograde flow (; mid row), but no correlation between retrograde flow and assembly was observed (; bottom row). This indicates that during a protrusion event, retrograde flow is transiently decreased, thereby supporting forward propulsion of the cell edge by growing F-actin. The modulation of retrograde flow at the leading edge is independent of F-actin assembly, and therefore unrelated to the assembly-mediated modulation of retrograde flow seen in leader cells.
Characterization of Lamellar Actomyosin Structures
We hypothesized that the less persistent protrusions of follower and island cells is driven by two weakly synchronized processes, F-actin assembly in the lamellipodium and variations in retrograde flow which depend on F-actin contraction of the lamella [31
]. On the other hand, these processes in leader cells may be mechanically isolated by adhesion formation near the cell edge. To test this possibility we characterized the organization of the contractile and adhesion systems in these three cell types by immunolocalization of F-actin, myosin II, and paxillin. In leader cells, long actomyosin bundles were organized parallel to the direction of migration. We refer to these structures as longitudinal bundles (; yellow regions) [32
]. Myosin II staining was absent from the utmost lamellipodia [5
] and gradually accumulated in the lamella co-localizing with F-actin bundles. Numerous paxillin-containing sites were present at the cell edge (; arrowhead) with an elongated appearance marking the ends of longitudinal bundles (; arrows). This suggests that adhesions are under posterior-anterior tension [34
]. In contrast, follower and the majority of island cells were characterized by thick bundles with dense myosin II staining arranged perpendicular to the direction of migration (; blue regions). We refer to these structures as transverse actomyosin arcs [35
]. Paxillin staining revealed numerous adhesion sites that appeared less regular in size and orientation than those observed in leaders (; arrowheads). Interestingly, a small minority of island cells resembled the morphology of leader cells with an expanded lamella, array of longitudinal bundles, and elongated adhesion sites (; arrowheads). These observations suggest that the organization of F-actin structures in the lamella and the dynamics of F-actin assembly and protrusion at the leading edge are tightly coupled and thus may be functionally related.
Immunolocalization of F-actin, myosin-II and adhesions in PtK1 cells
Characterization of Lamellar Contraction Modes
To further characterize the contractility of lamellar actomyosin structures we imaged and tracked F-actin speckle movements. Leader and island cells with high protrusion persistence displayed large areas of converging F-actin flow, with anterograde flow distal from the cell edge outpacing proximal retrograde flow (; Supplemental Data, Video 3
). Peaks in fluorescence intensity of eGFP-myosin II were observed in this convergence zone (; highlighted region) [36
]. Closer inspection of speckle motion by specialized single particle tracking methods (see Exp. Procedures) indicated that the apparent zero net speeds of F-actin speckles reported by coarser-grained flow analyses result from mutual cancellation of interdigitated anti-parallel movements of individual speckles. This is likely the product of thin F-actin fibers being cross-linked by myosin II [38
]. Conversely, non-persistently protruding cells displayed retrograde flow of F-actin throughout the narrow lamella, but lacked significant anterograde flow. Multi-wavelength time-lapse images of F-actin and eGFP-myosin II indicate that retrograde flow of actomyosin filaments were continuously compressed into a gradually denser network of transverse actomyosin arcs [38
](; Supplemental Data, Video 4
). Some follower and island cells exhibited a third contraction mode, in which retrograde and anterograde flows slide past one another in a shear-like fashion without interdigitated speckle movements (). From these experiments we conclude that persistent protruders show a strong degree of posterior - anterior connectivity established by contractile actomyosin structures that span the entire zone of the lamella and cell body, while non-persistent protruders lack this alignment of the contractile machinery with the direction of protrusion.
Myosin II-Inhibited Cells Exhibit Persistent Protrusions
Several reports have shown that inhibition of myosin II results in faster protrusion and motility rates [39
], suggesting a negative role of actomyosin contraction in promoting cell protrusion. Moreover, it was argued that decreased myosin II contractility activates Rac1 [40
] which may promote heightened F-actin assembly and leading to the enhanced protrusion rates. Both data seem inconsistent with our proposal that myosin II-mediated posterior - anterior connections may be required for persistent protrusion. To address this controversy we blocked myosin II function by blebbistatin [41
]. After two hours of treatment with 50 µM blebbistatin, all wound edge cells exhibited persistent protrusions (). Immunolocalization studies confirmed the disintegration of actomyosin structures (Supplemental Data, Figure S1A
]. Paxillin staining revealed that elongated or larger adhesion complexes were replaced by tiny adhesion sites at the utmost edge, which were sufficient to sustain cell-substrate contacts during cell migration (Supplemental Data, Figure S1A
; arrowheads) [25
]. Edge tracking revealed efficient protrusion activity for all cells at the wound edge (Supplemental Data Figure S1B
, left panel). Analysis of mean protrusion, protrusion persistence, and instantaneous edge velocity were statistically indistinguishable from leader cells (p > 0.3; ). Yet, as expected, myosin-inhibition abolished both retrograde and anterograde flows in the lamella, but preserved lamellipodial flow (Supplemental Data Figure S1B
, mid and right panels) [5
]. Thus, the apparently identical protrusion behavior of leader and myosin-disrupted cells originated from profoundly distinct F-actin dynamics.
Relationships between edge movement, lamellipodial F-actin dynamics, and lamellar contraction probed during the perturbation of myosin-II
In blebbistatin-treated cells, correlation analysis between edge protrusion and F-actin assembly showed a peak at −20 seconds (, top row), similar to leader cells (; top row). However, the correlation peak was decreased and barely significant at a 95% confidence level. Correlation between edge protrusion and F-actin retrograde flow revealed two significant extrema (, middle row): i) A negative peak at −50s indicates transient increases in flow after maximal edge protrusions. Together with the negative correlation between flow and F-actin turnover, this suggests that like in leader cells protrusion is limited by the force levels adhesion sites sustain against F-actin propulsion at the leading edge [28
]. ii) A positive correlation peak at +10~20s indicates that edge protrusion is further correlated with transient reduction of retrograde flow, a characteristic feature of follower and island cells. Since retrograde flow in blebbistatin-treated cells is independent of lamellar contraction, these flow reductions are likely associated with spontaneous increases in the coupling of lamellipodium F-actin to the substrate. Overall, the mechanism of protrusion in blebbistatin-treated cells relied weakly on F-actin assembly and on the contraction-independent modulation of retrograde flow, establishing a phenotype in between leader and follower cells. This seemed incompatible with the idea that inhibition of myosin II would result in heightened F-actin assembly.
Comparison of Relative F-actin Assembly Rates in Migrating Cells
To confirm this notion, we used quantitative fluorescent speckle microcopy to compare rates of F-actin assembly among different cell types (). Blebbistatin-treated cells had three-fold lower assembly rates when compared to leaders. Follower and island cells also exhibited higher assembly rates than drug-treated cells. Thus, the surprisingly similar protrusion persistence of leader and contraction-inhibited cells is ascribed to different states of F-actin dynamics at the leading edge. In leader cells, the F-actin network grows with an overall surplus over contraction-driven retrograde flow (). Protrusion rates during edge advancement are largely determined by F-actin assembly, while during phases of edge pausing or retraction, assembly rates temporarily fall below the rates of F-actin retrograde flow. In follower and island cells, rates of retrograde transport are similar to those in leader cells, but the levels of filament assembly are reduced so that the magnitudes of network growth and net retrograde flow are similar, thus explaining the frequent switching between protrusion and retraction. In contraction-inhibited cells, repressed assembly rates are sufficient to drive continued edge advancement as a result of low or no retrograde flow in the lamella.
Our study provides a detailed characterization of the intricate balance between F-actin assembly and retrograde flow that is required for persistent protrusion of epithelial cells during a wound healing response. The selection of this cell model allowed us to analyze the differences in the relations between assembly and flow that yield different protrusion persistence in leader and follower cells. Furthermore, these cells exhibit very heterogeneous protrusion behaviors along the cell edge and over the allotted observation time. We exploited this heterogeneity to identify locally the temporal coordination between assembly, flow and protrusion behaviors and to derive from it the requirements for cell edge movement. Together, these data defined precise functional linkages between F-actin dynamics and protrusion events.
Our data can be summarized in a network of synergistic and antagonistic mechanical cues (). The thickness of the connections indicates the importance of the relations between the cues. In leader cells protrusion is driven by high levels of F-actin assembly and nascent adhesion formation. The latter blocks the effects of strong lamellar contraction on retrograde flow, which otherwise would be a major inhibitor of protrusion. Transient increase in filament assembly and/or decrease in adhesion formation generate moderate increases in retrograde flow that temporarily reduce protrusion efficiency. In follower and island cells the levels of filament assembly and adhesion formation are lower, resulting in higher influence of lamellar contraction on retrograde flow. In these cells the edge advances only sporadically when transient increases in adhesion formation generate moderate decreases in retrograde flow, sufficient for the filament assembly to temporarily propel the edge forward. As soon as the adhesions get weaker, higher rates of retrograde flow cause renewed retraction of the cell edge. Hence, these cells exhibit a characteristic instability in protrusions with very low net rates of edge advancement.
Integrated model of cell protrusion
In blebbistatin-treated cells the levels of filament assembly and adhesion formation overall are also much lower than in leader cells. However, due to the absence of lamellar contraction, retrograde flow is concentrated to the lamellipodium and reduced to a level where low rates of filament assembly are sufficient to drive protrusion at the same magnitude as in leader cells. Sporadic flow increases, accompanied by protrusion decreases, are observed when the rate of adhesion assembly transiently decreases and/or the rate of filament assembly increases. Therefore, these cells exhibit positive correlation between filament assembly and protrusion and delayed negative correlation between retrograde flow and protrusion like in leader cells, as well as showing correlation between retrograde flow and protrusion like in follower cells.
Our model makes assumptions for two additional interactions. First, it posits that lamellar contraction promotes the formation of nascent adhesions via a mechanical signal. Thus, the contraction machinery positively influences the rates of protrusion. Second, the model posits that nascent adhesions are hubs for mechanotransduction, where contractile forces and forces balancing assembly-mediated propulsion are converted into chemical signals, which in turn control again the rates of filament assembly and adhesion formation.
Evidence for the assumption of a contraction-dependent formation of nascent adhesions comes from both the literature and our own experiments. Positive cross-talk between contractile forces and adhesion formation has been a long-established phenomenon [34
]. Recently more focused data have emerged that explain the force-sensitive assembly of adhesion complexes on a mechanistic level [44
]. Our own data provides several correlative lines of evidence for such cross-talk. Paxillin-staining indicated distinct adhesion morphology and distribution for leader, follower and contraction-inhibited cells ( and Fig. S1
). In leader cells, numerous elongated adhesions align with the direction of migration. Co-staining of F-actin and paxillin indicates that these adhesions are directly linked via thin actin fibers to the convergence zone at the base of the lamella. Consistent with observations by Riveline et al.,[48
] and our analysis of anti-parallel speckle flow in the convergence zone and lamella, we suggest that the formation of these adhesions at the cell edge are mediated by long-range anterior-posterior transduction of contraction forces to the cell edge via longitudinal bundles. In follower cells, where longitudinal bundles are largely absent, paxillin staining displays much less polarity in the adhesions. Thus, we propose that force transduction to the cell edge is less efficient and less directed, preventing a persistent, force-induced assembly of adhesions that counteract retrograde flow. This model is corroborated by the observation that blebbistatin-treated cells exhibit small and non-polarized adhesions (Supplemental Data, Figure S1A
Evidence for links between force-stimulated adhesion formation and chemical signaling has been established by a large body of work, especially for signals related to the activation of the family of small RhoGTPases [46
]. The mechanisms of such mechanotransduction are still largely elusive; but, it is clear that both Rac1 and RhoA are activated at the leading edge [51
], and it is conceivable that these signals feedback into pathways that regulate adhesion formation [53
]. Both signaling molecules are also upstream of F-actin nucleators [55
]. Relations between contractility and RhoGTPase signaling have also been described. Most prominently, recent evidence has shown that inhibition of actomyosin contraction leads to increased Rac1 activation [40
]. However, these relationships have been established only on a global scale. To determine the coupling between contraction forces and signaling at the spatial and temporal scales relevant to the regulation of cell protrusion, it will be necessary to simultaneously measure contraction events and signal activation at second and sub-micron resolutions. Our model would predict that signaling activities are selectively elevated after contraction events. While the basic technologies for the analysis of such correlations are now available [52
] significant experimental hurdles have to be overcome to monitor contraction forces and signaling at the required resolution. Alternatively, high-resolution laser ablation technology is becoming available with which it should be possible to disrupt selectively longitudinal F-actin bundles that connect contraction events to adhesions. Based on our model, significant reduction in RhoGTPase signaling is expected at adhesion sites in which cut F-actin bundles terminate.
Our current data provide indirect evidence for a link between contraction and signaling. The front-coupled contraction forces in leader cells are accompanied by high rates of filament assembly, which likely requires elevated signaling activity at the cell edge. The less-directed transduction of contraction forces to adhesions in follower cells is accompanied by lower rates of F-actin assembly. Finally, abrogation of contraction by blebbistatin treatment drops the assembly rate approximately threefold. Therefore, we propose that the level of lamellar contraction and the structural efficiency of posterior-anterior force transduction to adhesions at the cell front critically determine the level of F-actin assembly in the lamellipodium.
Our observation of reduced F-actin assembly after blebbistatin treatment seems to contradict the notions of elevated Rac1 activity under blocked myosin II contraction [40
] and of Rac1 being a promoter of F-actin polymerization [57
]. We interpret our data as an indication that in the absence of contraction, Rac1 may inhibit a Rac1-independent pathway of F-actin nucleation. Given some biochemical evidence that Rac1 antagonizes RhoA, [58
] it is tempting to speculate that the elevated Rac1 levels oppress a RhoA-mediated assembly pathway. Whether such a Rac1-independent assembly pathway significantly contributes to protrusion in leader cells with an intact contractile system, and whether intrinsic variations in the contractile activity yield temporally alternating levels of contribution between Rac1-dependent and Rac1-independent assembly pathways, again will require concurrent measurements of actin assembly rates and signaling activities at the scale of protrusion regulation.
In conclusion, this study illuminates via local measurements of F-actin dynamics the subtleties of a spatiotemporal balance between lamellipodial F-actin assembly and the dichotomy of pro- and contra-protrusion cues mediated by different actomyosin contractile structures in the lamella. It sets the stage for future analyses where the interdependences of contraction, assembly, and signaling events are resolved by simultaneous recording of these parameters across space and time.