Maturation of Filopodia Shaft Adhesions Is Upregulated by Local Cycles of Lamellipodia Advancements and Retractions

While cell-substrate adhesions that form between the protruding edge of a spreading cell and flat surfaces have been studied extensively, processes that regulate the maturation of filopodia adhesions are far less characterized. Since little is known about how the kinetics of formation or disassembly of filopodia adhesions is regulated upon integration into the lamellum, a kinetic analysis of the formation and disassembly of filopodia adhesions was conducted at the leading edge of β3-integrin-EGFP-expressing rat embryonic fibroblasts spreading on fibronectin-coated glass or on soft polyacrylamide gels. Filopodia β3-integrin adhesions matured only if the lamellipodium in their immediate vicinity showed cyclic protrusions and retractions. Filopodia β3-integrin shaft adhesions elongated rapidly when they were overrun by the advancing lamellipodium. Subsequently and once the lamellipodium stopped its advancement at the distal end of the filopodia β3-integrin adhesion, these β3-integrin shaft adhesions started to grow sidewise and colocalize with the newly assembled circumferential actin stress fibers. In contrast, the suppression of the cyclic protrusions and retractions of the lamellipodium by blocking myosin light chain kinase suppressed the growth of filopodia adhesion and resulted in the premature disassembly of filopodia adhesions. The same failure to stabilize those adhesions was found for the advancing lamellipodium that rapidly overran filopodia shaft adhesions without pausing as seen often during fast cell spreading. In turn, plating cells on soft polyacrylamide gels resulted in a reduction of lamellipodia activity, which was partially restored locally by the presence of filopodia adhesions. Thus filopodia adhesions could also mature and be integrated into the lamellum for fibroblasts on soft polyacrylamide substrates.

Although never addressed in the case of filopodia shaft adhesions, it has been demonstrated that the maturation of focal adhesions in the lamellum, is a myosin II-dependent process [35], but not that of nascent adhesions in the lamellipodium [36], and that focal adhesion maturation is coupled to the cycles of lamellipodium protrusions and retractions [37]. The lamellipodium exhibits cycles of protrusions and retractions, whereby the protrusions are driven by the actin polymerization of the dendritic actin network at the leading edge of the lamellipodium [38] while the retractions are initiated by actin-myosin II located in the lamellum just behind the lamellipodium-lamellum transition zone [35,37,39]. The formation and dynamics of filopodia and lamellipodia are regulated by different GTPases [40][41][42][43][44], suggesting that the formation, maturation and turnover of the adhesions associated with these two types of cell edge protrusions serve different purposes.
In fish fibroblasts, the appearance of filopodia adhesions, as recorded by paxillin recruitment, was observed to coincide with the advancement of the lamellipodium up to and past the respective adhesive segment of the filopodium [21]. In keratinocytes, when reached by the advancing lamellipodia, nascent filopodia adhesions, visualized via different cytoplasmic adaptors and proximity to the substrate, increase their size along the former orientation of filopodia resulting in more matured substrate adhesions [22], a maturation process which has been proposed to be associated with increased traction forces [32].
To understand the kinetic and mechanical aspects of how filopodia shaft adhesions grow and mature, and how this synchronizes with the protrusion of lamellipodia, we decided to visualize the formation of filopodia adhesions with respect to the clustering of b3-integrin receptors in the filopodia membrane. The b3-integrin is one of the major transmembrane proteins in the formation of earliest cell-substrate adhesions [28,45]. While a5b1 integrins specifically recognize fibronectin (FN) and the clustering of a5b1 integrins determines the strength of cell-substrate adhesions [46], avb3 integrins mediate the rigidity sensing at the cell leading edge [47] and, together with talin, enable early mechanotransduction events [46]. To better understand how filopodia shaft adhesions are formed and get incorporated in the lamellum, we monitored the substrate adhesion and spreading of rat embryonic fibroblasts (REF52) that were stably transfected with b3-integrin-EGFP. This b3-integrin-EGFP construct has previously been validated carefully and utilized to study the kinetics of integrin clustering [45,48,49], as well as for studies detailing the mechanism of how the interaction of talin with b3integrin enables talin activation and reciprocal integrin clustering [50]. Here we were particularly interested in characterizing the growth dynamics of filopodia shaft b3-integrin clusters in response to the local dynamics of protrusion and retraction of lamellipodia. In addition, the impact of lamellipodia dynamics on the characteristics of cytoskeleton organization at filopodia adhesion was examined. Since the cycles of lamellipodium protrusion and retraction are more pronounced on rigid than on soft polyacrylamide gels coated with FN, we finally asked how these different cyclic motions of lamellipodia due to substrate rigidity might affect the life cycle of filopodia shaft adhesions.

Results
Elongation and growth of b3-integrin-EGFP filopodia shaft adhesions is promoted in proximity to an advancing lamellipodium To capture and quantify the kinetics by which filopodia shaft adhesions form, mature and get incorporated in the lamellum, REF52 fibroblasts stably transfected with b3-integrin-EGFP (REF52-b3-integrin-EGFP) were allowed to adhere and spread on fibronectin coated glass slides. By quantitative analysis (see methods) of time-lapse confocal fluorescence microscope sequences, we tracked the growth kinetics of b3-integrin filopodia adhesions. Several sequential growth phases of filopodia adhesion could be distinguished, providing information on how filopodia adhesions grow and get integrated into the cell lamellum ( Fig. 1, Video S1).
Phase 1: Filopodia b3-integrin shaft adhesions rapidly elongated rearward towards the cell edge (Fig. 1A, a-d; 1B, white zone). Shortly after the rapid protrusion of a filopodium (,80 nm/s for ,1 min ( Fig. 1A-a, b)), we often observed the formation of a first surface contact as indicated by the occurrence of a green integrin cluster close to the filopodium tip ( Fig. 1A-c). Immediately thereafter, the lamellipodium at the filopodium base started to advance along the filopodium shaft towards this filopodia integrin cluster (Fig. 1A, c-d). During the advancement of the lamellipodium, the length of the filopodium integrin shaft adhesion increased rapidly (Fig. 1B, red curve in white zone), while the width of the integrin adhesion remained unchanged (Fig. 1B, blue curve in white zone). The resulting rapid increase of the aspect ratio of the filopodia shaft adhesion is given as well (Fig. 1B, grey curve in white zone).
Phase 2: Rearward growing filopodia b3-integrin shaft adhesions are often overrun by advancing lamellipodia ( Fig. 1A-d, e; 1B, yellow zone). For the filopodium shown, the advancing lamellipodium reached the proximal end of the filopodia integrin adhesion 120 s after the filopodia adhesion was initiated (Fig. 1A-d). It took about additional 80 s for the lamellipodium to advance across the filopodium shaft adhesion and reach its distal end (Fig. 1A-e). This resulted in the full entry of the filopodium adhesion into the lamellipodium. In this period, the depicted filopodium adhesion increased its width, while the length further increased reaching its maximal length.
Phase 3: Integration of the filopodia b3-integrin shaft adhesions into pausing lamellipodia ( Fig. 1A-e, f, g and their overlays; Fig. 1B, pink zone). After reaching the distal end of a filopodium b3-integrin shaft cluster, the lamellipodium typically paused at this position without net advancement for the rest of the tracked duration. Correspondingly, the filopodia b3integrin adhesions that integrated into lamellipodia did not grow further in length, but continued to widen, causing a rapid decrease in their aspect ratio while their total area further increased. Here after 600 s of growth, the depicted filopodium shaft adhesion had significantly increased in width and area. New b3-integrin clusters (Fig. 1A, white arrows) also started to form at their sides and eventually merged with the sidewise growing filopodia adhesions.
While the above analysis was based on the b3-integrin adhesions initiated at the distal tip of the filopodium (filopodia tip adhesions), similar b3-integrin adhesions could also be formed at the bases of filopodia (filopodia base adhesions), which were in contact with lamellipodia from the beginning of their initiation ( Fig. S1; Video S2). Importantly, comparable sequences of events were observed for the maturation of filopodia shaft adhesions in REF52 and human foreskin fibroblast (HFF) cells (Figs 1, 2A-C, S1A-C, S4A,B). Upon maturation, filopodia adhesions generally existed for more than 10 minutes ( Fig. 2A-C), with some over 20 minutes (Fig. S1A, B). Furthermore, in cases where the filopodia had a non-adherent distal section, these filopodia tips deteriorated (bending, fracturing, recycling or lateral movement (Figs S1A, S4A; Video S2)) as observed previously in fish fibroblasts [21]. However, these movements of non-adherent distal filopodia tip segments did not alter the growth characteristics of more proximal filopodia adhesions (Fig. S1A, B, black arrows). Overall, the growth characteristics of filopodia adhesions were distinctively different from point-like nascent adhesions formed at the lamellipodium type of cell edge ( [36], Fig. S5).

Time-dependent quantification of b3-integrin-EGFP intensity and vinculin recruitment
To quantify the fluorescence intensity of filopodia b3-integrin-EGFP clusters as a function of time (Fig. 2D), the average fluorescence intensity per pixel of filopodia b3-integrin-EGFP clusters was measured. It rapidly increased when first contacting the advancing and later the pausing lamellipodium (Fig. 2D, yellow and pink segments respectively). Shortly after the onset of the lamellipodial pausing, the fluorescence intensity reached a plateau, and then slightly declined due to photobleaching.
To quantify the impact of the pausing lamellipodium, we analyzed the spatial densities of b3-integrins in filopodia adhesions as probed by the fluorescence intensity profile (Fig. 2E) along the long axis of the filopodia b3-integrin-EGFP cluster (Fig. 2E, insert) at the end of its tracked duration. For the steadily maturing filopodia adhesions, the fluorescence intensity of b3-integrin-EGFP clusters peaked at a distance of 0.960.6 mm (n = 8, 4 cells) away from the distal ends of filopodia adhesions, and then decreased towards their proximal ends.
Since vinculin recruitment to newly formed integrin adhesions requires force-activation of talin or a-actinin [26,27,30,31,[51][52][53], with a close correlation of force and vinculin recruitment to focal A. Selected frames from a confocal microscopy time-lapse sequence of a growing b3-integrin-EGFP cluster (green) (44 min-61 min after plating the cell, Video S1). The cell membrane was visualized by the membrane stain DiI (red). The dashed rectangle (top of the right cell image) was magnified in the left column (a-g). A rapidly protruding filopodium (a-b) forms a surface contact with the filopodium tip as seen by the initiation of b3-integrin-EGFP cluster (c). The white dashed vertical line indicates the initial tip position of the filopodia adhesion site. The colored frames in A correspond to the colored phases as classified in B, i.e. the time periods in which the b3-integrin-EGFP cluster was sequentially in contact with the advancing lamellipodium (yellow) and with the pausing lamellipodium without net advancement (pink) respectively. On the right bottom, the b3-integrin-EGFP images from different time points were overlaid (cyan-early, white-later). White arrows point to the newly formed b3-integrin-EGFP cluster at the side of the filopodia adhesion. B. Quantification of the growth kinetics of the filopodia b3-integrin-EGFP cluster and the associated filopodium. The letters (a-g) marked the time of the corresponding images (A, a-g). To indicate the movements of the lamellipodium when it is in contact with the growing filopodia b3-integrin adhesion, the plots were color-coded into zones corresponding to the color framed images as defined in A. The plots present the quantification of the filopodia b3-integrin-EGFP cluster characteristics, including its width (blue), length (red), area (black) and aspect ratio (grey).  , area (C) and average fluorescence intensity (D, average = sum of grey value/number of pixels) were plotted for b3-integrin-EGFP clusters for filopodia tip adhesions (thin curves, representing 4 maturing filopodia adhesions in 3 cells) and filopodia base adhesions (thick opaque curves, representing 2 maturing filopodia adhesions in 1 cell). Curves Tip-a and Basea corresponded to the filopodia adhesions shown in Fig. 1 and Fig. S1 respectively. The traces were color coded (similar to the colored zones in Fig.  1B) to indicate the different local movements (advancing or pausing) of the lamellipodium in relation to the growth of the filopodia adhesion (black: before being reached by lamellipodium; yellow: being passed by an advancing lamellipodium; pink: with the lamellipodium persisted at the distal end of the filopodia adhesion without net advancement). E. The spatial distribution of b3-integrins in filopodia adhesions. For the filopodia adhesions analyzed in A-D, the fluorescence intensity profiles were generated along the long axes of filopodia b3-integrin-EGFP clusters in the last frame of their tracked sequences, as indicated by the dashed arrow in the Insert (same as analyzed in Fig. 1; green: b3-integrin-EGFP; red: membrane). The range of the x axis for the intensity profile curves corresponded to the full lengths of filopodia adhesions between their distal and proximal ends. adhesions in live HFF cells [54], we immunostained for vinculin in HFF cells 20 min after cell seeding (Fig. 2F). In agreement with the literature [20,22], we see that vinculin gets recruited not only to focal adhesions, but importantly also to the filopodia shaft adhesions already prior to their integration into the lamellum (Fig. 2F). Here we find that vinculin recruitment to filopodia shaft adhesions was upregulated once they got in contact with the lamellipodium (Fig. 2F). The average (per pixel) intensities, as well as the area, of recruited vinculin in the filopodia adhesion in contact with a lamellipodium (Fig. 2F, pink arrowheads) were significantly higher than those of the filopodia shaft adhesions located in front of a lamellipodium (Fig. 2F, grey arrowheads, Fig. 2G). In addition, the vinculin staining fully colocalized with the dark regions in the interference reflection image (Fig. 2F-c, f), indicating a tight surface anchorage of these filopodia adhesions to the substrate. This suggests that tensile forces are already applied to the filopodia shaft adhesions, and that the force-activated vinculin recruitment is upregulated once filopodia shaft adhesions get integrated into the lamellipodium.

Lamellipodia undergo cyclic protrusions and retractions in the proximity of filopodia adhesions
Previous studies of spreading cells [14,35,37,55] have shown that lamellipodia undergo cycle of protrusion and retraction in fibroblasts adhering to rigid surfaces [37]. To examine the dynamics of lamellipodia at filopodia adhesions, we analyzed the confocal fluorescence time-lapse data of the REF52-b3-integrin-EGFP cells plated on FN coated glass (Fig. 3). The dynamics of the lamellipodium movement were monitored by DiI fluorescent staining of the cell lipid membrane (red) with respect to the filopodia b3-integrin adhesion (green). As demonstrated by the fluorescence montage of the cell edges, periodic lamellipodia protrusions and retractions occurred in proximity to filopodia b3integrin adhesions even after their integration into the lamellipodium was completed (Fig. 3C, Video S3). To better visualize the protrusions and retractions of lamellipodium in the vicinity of a filopodia adhesion, we generated kymographs at multiple positions along the cell edge close to the filopodia b3-integrin adhesion (Fig.  S2A). Comparison of corresponding kymographs demonstrated that the spatio-temporal patterns of the periodic protrusion and retraction cycles of the lamellipodium remained similar in the vicinity of filopodia b3-integrin adhesions (18 filopodia adhesions in 6 cells, see e.g. Fig. S2A-b), and that the kinetics are similar for filopodia-containing cell edges in both spreading REF52 cells and HFF cells on FN coated glass (Fig. S2B, D). For REF52 cells, the leading edge protruded on average for 38630 s and then retracted for 28621 s., while the average distances traveled by lamellipodia during their protrusions and retractions were 1.360.8 mm and 1.260.8 mm respectively (Fig. 3D). It should be noted that these similarities between REF52 and HFF cells are expected since the molecular modules used by different fibroblasts and by many other cells, by which integrins are coupled to the actin cytoskeleton, are the same.

Inhibition of myosin light chain kinase suppresses the growth of the filopodia shaft adhesions
Myosin II is concentrated behind the lamellipodium/lamellum transition zone [56] and the activation of myosin II by myosin light chain kinase (MLCK) was shown to be necessary for the periodic contractions of lamellipodia [35,37]. We thus biochemically inhibited MLCK in spreading REF52-b3-integrin-EGFP cells using the specific inhibitor ML-7. As predicted, this treatment inhibited the cyclic protrusions and retractions of the lamellipodia (5 s/frame) (Fig. 4A, Video S4), and the edges of ML-7 treated cells were dominated by lamellipodia ( Fig. 4B-a). Far fewer filopodia were seen in ML-7 treated cells and the few that formed had filopodia b3-integrin adhesions of considerably suppressed growth (n = 16) ( Fig. 4B-b), and a significantly smaller size (0.6160.25 mm 2 , p,0.001) when compared to maturing filopodia b3-integrin adhesions in untreated cells (1.3760.95 mm 2 , Fig. 4Bc). Therefore, our data suggested that the acto-myosin activity is required for the maturation of filopodia shaft adhesions.
Filopodia adhesions stabilize lamellipodia and thereby enable their local cyclic protrusions and retractions on soft polyacrylamide gel Since cyclic protrusions and retractions of lamellipodia were previously shown to occur only on rigid, but not on FN-coated soft polyacrylamide (PAA) gels [37] where the cell adhesions were found to be more diffuse and unstable [57], we studied the spreading of REF52-b3-integrin-EGFP cells on soft (7.4 kPa) PAA gels covalently coated with FN. Due to the lack of stable adhesion of the cell edge to soft substrate, the lamellipodium along the cell periphery demonstrated significant ruffling ( Fig. 4C-b, Video S5), which is in agreement with the previous literature [37]. However, filopodia shaft adhesions could form on soft PAA substrates and were integrated into lamellipodia, which coincided with a reduction in lamellipodial ruffling activity (

Reduced contact time between filopodia shaft adhesions and advancing lamellipodia blocks their maturation
Though less often observed at our highest spatiotemporal resolution, some filopodia shaft adhesions (,10%) rapidly disassembled after their formations. One of such filopodia b3integrin shaft adhesions (Fig. 5A, Video S7) was located at a short filopodium (,2 mm in length). Typically, it took less than 20 s (yellow horizontal bar) for the advancing lamellipodium to move across such filopodia shaft adhesions. After briefly pausing (,60 s, pink horizontal bar) at its distal end, lamellipodia resumed their advancement and left the filopodia adhesion behind in the cell lamellum (grey horizontal bar). The b3-integrin clusters of these filopodia adhesions were small and disassembled rapidly. This was in drastic contrast to the long stability of maturing filopodia shaft adhesions that were in contact with a pausing lamellipodium as analyzed earlier (Fig. 1).
The failed growth of these filopodia adhesions was further quantified by monitoring the growth traces over time with respect to the area, length, width, and average fluorescence intensity of their b3-integrin-EGFP clusters (Fig. 5B). Although they started with similar sizes as compared to the later maturing filopodia shaft adhesions ( Fig. 2A-C), their growth was suppressed (trace 3) or they rapidly disintegrated (traces 1 and 2) ( Fig. 5B-a, b, c). Similarly as compared to the maturing filopodia adhesions (Fig. 2D), the fluorescence intensities of the disassembling filopodia b3-integrin-EGFP clusters had temporally increased during their brief contact with lamellipodia ( Fig. 5B-d, yellow and pink segments), but they rapidly decreased after the separation of the advancing lamellipodium from these filopodia adhesions. Importantly, all the observed unstable filopodia adhesions had very short total contact time with lamellipodia (Fig. 5B, yellow and pink segments), with an average duration of 69647 s (Fig. 5C). This is significantly shorter than observed for the maturing filopodia adhesions (5426360 s, limited by the tracking durations, Fig. 5C), which stayed in contact with a pausing lamellipodia. Sufficient contact time with the local pausing lamellipodium thus seem to be critical parameter regulating the maturation of filopodia adhesions.
Slow lamellipodium advancement increased the contact time and promoted the maturation of filopodia adhesions Next, we quantified the net advancement speed of the cell edge. This speed at the filopodia adhesion was determined as Dd/Dt (net advancement distance of cell edge (Dd) divided by the total time of cell edge movement (Dt)) in the kymograph (Fig. S2C). The part of the contact time between filopodia adhesions and the advancing lamellipodia (e.g. Fig. 1, yellow zone) was plotted as function of this speed of the spreading REF52-b3-integrin-EGFP cells (Fig. 6A). Our data demonstrated a wide range of contact times (10-400 s) of filopodia adhesions with the advancing lamellipodia. The net advancement speed of the cell edge was below 40 nm/s for most of filopodia adhesions whereby the lamellipodia in all of these cases first advanced and then paused for long durations at the distal ends of the filopodia adhesions which steadily matured. Similarly in spreading HFF cells, we could observe a similarly wide range of cell edge advancement speeds (9-100 nm/s) and the long duration of positional pausing of the slowly advancing lamellipodia at the distal tips of filopodia adhesions (Fig. S4).
In summary, lamellipodia were often found to either pause at the distal ends of the filopodia shaft adhesions or to keep advancing (Fig. 6A, black and red data points respectively). This resulted either in their subsequent maturation or disassembly respectively, when the cell edge advanced too rapidly (.40 nm/s). These two modes of contact between the fast advancing  Circumferentially oriented actin fibers intersect with former filopodia shaft adhesions within the cell lamella Finally, to examine the actin cytoskeleton organization at filopodia adhesions, we used a newly developed ventral cell membrane preparation method which allowed us to expose the partially well-preserved ventral cell membrane with its associated cell-substrate adhesions and actin structures. During the cell burst, only those actin filaments that are firmly anchored to the ventral membrane remain attached. Figure 7 shows that, in adherent REF52-b3-integrin-EGFP and HFF cells, the former filopodia shaft adhesions (Fig. 7, pink arrows and arrowheads) were connected to the circumferentially (in parallel to the cell edge) oriented actin stress fibers of varied thickness (Fig. 7, cyan  arrowheads). These fibers were located in the ventral cell membranes, and often in tight contact with the substrate surface. Filopodia adhesions also retained the radial (toward the cell center) oriented original filopodia actin bundles located in ventral cell membranes (Fig. 7, actin images, pink arrows and arrowheads). While the newly formed filopodia adhesions fully colocalized with the slender filopodia actin bundles (Fig. 7A,C, pink arrows), more developed filopodia adhesions with sidewise widened b3-integrin clusters were associated with the sidewise deformed filopodia actin bundles (Fig. 7B,C, pink arrowheads).

Discussion
Here we describe for the first time the kinetics how the maturation of filopodia shaft adhesions correlates with the cyclic protrusion and retraction activities of the lamellipodium. While filopodia adhesions were previously found to be integrated into the cell lamella of various cell types [20][21][22]32,58], no information was available about the integration process. Here we show that filopodia shaft adhesions grow and mature in sequential phases that correlate with the advancing and pausing of the adjacent lamellipodium (Figs 1, 2, S1). We also show that vinculin gets recruited into the filopodia shaft adhesions (Figs 2F, 7A), which suggests that they are exposed to tensile forces. In contrast to the previously described behavior of nascent adhesions within lamellipodia [36], which remain point-like and can be turnedover rapidly [36], the b3-integrin-rich filopodia shaft adhesions once formed increase rapidly in length and their initial formation typically induces the advancement of the proximal lamellipodium (phase 1) (Figs 1, 2A, S1; Video S1). In the next phase, the rearward growing filopodia shaft adhesions are overrun by the advancing lamellipodium (phase 2) until the point, where the lamellipodia pauses at the distal edges of filopodia shaft adhesions (Figs. 1, S1; Video S1, S2). Once lamellipodia started to pause, the now integrated filopodia shaft adhesions started to grow in width (phase 3) (Figs 1, 2A, 2B, S1).
Previous studies have shown that the activity of myosin II just behind the lamellipodium-lamellum transition zone regulate the cyclic retraction of the lamellipodium [24,35,37,59] and the maturation of cell-matrix adhesions and substrate rigidity sensing [35,37]. Our work could show now that the maturation of filopodia adhesions and their integration into the lamellum are also regulated by tensile forces. This was suggested by the upregulated recruitment of vinculin (Fig. 2F,G), which only binds to stretched but not relaxed talin [26,27], and the temporal increase of integrin fluorescence in filopodia shaft adhesions (Fig. 2D) once reached by the advancing lamellipodium, and by the existence of the cyclic protrusions and retractions of lamellipodia in the proximity of filopodia adhesions on FN-coated rigid (Figs 3A, S2A) and soft (PAA, Fig. 4D) surfaces. MLCK inhibition blocked the periodic protrusions and retractions of the lamellipodium (Fig. 4A,B) as previously described [37]. But MLCK inhibition did not perturb the initial formation of filopodia adhesions and their subsequent incorporation into lamellipodia, probably because the leading edges of cells are pushed forward by the protrusive forces generated by the barbed end elongation of the dendritic actin network [38]. Interestingly, the spatial distribution of the fluorescence intensity in the mature filopodia b3-integrin-EGFP cluster typically revealed a characteristic intensity maximum at about 1-1.5 mm away from the distal tip of the filopodia adhesion (Fig. 2E). This distance is similar to the often described distance from the cell edge to the lamellipodiumlamellum transition [29,[60][61][62]. These further suggested the maturation of the filopodia b3-integrin-EGFP cluster is coupled to the cycles of myosin II-generated forces.
The stability and lateral growth of b3-integrin-EGFP clusters originating from filopodia shaft adhesions (Figs 2A-C, 5B, S3) are shown here to be dependent on the contact time with the periodically advancing and retracting lamellipodium (Fig. 5C). Filopodia shaft adhesions mature upon their integration, but only when the local cyclically retracting lamellipodium paused its net advancement for a long duration at their distal tips (Figs 1, 2, 3, S1, S3). In contrast, the disassembled filopodia adhesions exhibited only short contact time with lamellipodia (69647 s, Fig. 5C), which occurred often when lamellipodia edge rapidly passed filopodia adhesions (Figs 5; 6B-b, S4C). Alternatively, when the cyclic protrusions and retractions of lamellipodia did not occur (Fig. 4B), filopodia adhesions failed to mature. Particularly in the former case, the contact time between lamellipodia and the disassembling filopodia adhesions (Fig. 5C) is similar to the time period of the lamellipodia retraction cycles (Fig. 3D), suggesting that maturation or reinforcement did not take place during the lamellipodial advancement phase. Thus the contact time and multiplicity of retraction cycles of the lamellipodium at the filopodia adhesions regulates the maturation or disassembly of filopodia adhesions. Cyclic protrusions and retractions of lamellipodia have also been associated with the deposition of circumferential actin filaments that run parallel to the cell edge in various cell types [35,55,63] which became efficiently coupled with focal adhesions at the onset of the lamellum right behind the transition zone [64]. Our data now suggest that the circumferential actin filaments become connected with filopodia adhesions behind the lamellipodium-lamellum transition (Fig. 7).
To conclude, filopodia and lamellipodia both sense the mechanical aspects of a cell's environment and their respective activities are key in the cell migratory behaviors, such as cell spreading [2,7,8,14,23], the path-finding of neuronal growth cones [65,66], angiogenesis [12], as well as cancer cell migration and invasion [13]. Our work now highlights the previously unrecognized interplay between the maturation of filopodia shaft adhesions and their regulation by the cyclic advancing and retraction dynamics of the proximal lamellipodium.

Substrate preparation
FN was isolated from human plasma (Zürcher Blutspendedienst SRK) using gelatin-sepharose chromatography based on established methods [67]. The Lab-Tek 4 wells chambered coverglasses (Thermo Scientific) were coated by adding 250 ml of 10 mg/ml FN in phosphate-buffered saline (PBS) and incubated for 1 hour at room temperature. Afterwards, the culture surface was removed of the FN solution and used for cell culture.
The polyacrylamide substrates with surface crosslinked FN were prepared following the procedure described by Tse and Engler [68]. In short, gels of approximately 7.4 kPa (or 35 kPa) stiffness and ,100 mm thickness were prepared by mixing 10% acrylamide (BioRad) and 0.06% (or 0.3%) bisacrylamide (BioRad) with distilled water, adding 1% APS (Sigma) and 0.2% TEMED (Sigma), mixing, and sandwiching 60 ml of this solution between a glutaraldehyde (Sigma) activated coverslip (25 mm diameter) and a dichlorodimethylsilane (Sigma) passivated coverslip (15 mm diameter). After removal of the passivated coverslip, gels were rinsed with distilled water and activated by adding 100 mL NHSdiazirine (Sulfo-LC-SDA, 1 mg/ml in 50 mM HEPES pH 8.2, Pierce Biotechnology) and illuminating for 4 min under a strong 365 nm UV source. This step was repeated for a total of 3 times. Then, gels were quickly rinsed 26 with distilled water and 150 ml of a 50ug/ml solution of FN in 50 mM HEPES pH 8.2 were Time-lapse montage of the cell edge region containing a disassembling filopodia adhesion in a b3-integrin-EGFP (green) expressing REF52 fibroblast on FN coated glass (40 min-53 min after plating, Video S7). This dissembling filopodia adhesion was located in the same cell region and captured in the same image sequence as for the maturing filopodia adhesion analyzed in Fig. 1. The cell membrane was stained by the fluorophore DiI (red). The colored bars indicate the relation between the filopodia adhesion and the lamellipodium as in Fig. 2A. Black segment: filopodia adhesion before being reached by the advancing lamellipodium. Yellow segment: passing of the advancing lamellipodium. Pink segment: pausing of the lamellipodium at the distal end of the filopodia adhesion without net advancement. Grey segment: the disassembly of the filopodia adhesion inside the cell lamellum after its full separation from the advancing lamellipodium. B. Time traces of the length (a), width (b), area (c) and average fluorescence intensity (d) of the disassembling filopodia b3-integrin clusters (same color code as for the maturing filopodia adhesions in Fig. 2A-D). Growth trace 1 corresponds to the dissembling filopodia adhesion in A. C. Comparison of the total contact time with lamellipodia (the averaged sum of the advancing and pausing durations of lamellipodia at filopodia adhesions) between the disassembling filopodia adhesions (n = 6) and those matured into stable adhesions (n = 18). Error bars correspond to the standard errors. doi:10.1371/journal.pone.0107097.g005 added and let react for 2 hours at room temperature. Gels were washed with PBS and sterilized by UV for 40 min in the cell culture hood.

Ventral cell membrane preparation
Fibroblasts cultured in the Lab-Tek 4 wells chambered coverglasses were extracellularly crosslinked with 250 ml solutions of Bis(NHS)PEO 5 (Pierce) (0.1 mM in PBS, 20 minutes) at room temperature. This solution was then exchanged to 500 ml of nanopure water to incubate the sample for 10 minutes. Afterwards, the sample was rinsed with 250 ml of PBS three times and then stained for observation. This ventral cell membrane method well preserves the ventral cell membranes and their associated substrate adhesions and actin structures, as compared to the commonly used whole cell sample preparations, or other procedures for the ventral cell membrane preparation, such as lysis squirting [69][70][71] or sonication [72]. The removal of dorsal cell membrane excludes the overlapping signals from the dorsal membrane structures, thus enables an unequivocal identification of actual structural connections of the ventral actin structures with cell-substrate adhesions. In order to avoid the alternation to the ventral membrane structures, samples were not further fixed and were examined immediately after their preparation.

Fluorescence staining
Cell membrane staining. Suspensions of the trypsinized cells were centrifuged (1200 rpm, 3 min, room temperature) and removed of the supernatant. The remaining cell pellet was resuspended in 1 ml of cell culture media containing the lipophilic membrane stain DiI (Invitrogen, 1:250 dilution), which does not introduce noticeable changes to the cell behaviors. After incubation at 37uC for 10 min, the cell suspension was centrifuged (1200 rpm, 3 min, room temperature) and removed of the membrane dye containing supernatant. The remaining pellets of the membrane stained cells were resuspended in 1 ml of cell culture media and used for experiments.
Immunofluorescence staining. Lysed cells or ventral cell membrane samples in Lab-Tek 4 wells chambered coverglasses were removed of the original buffer solutions, gently rinsed three times with 250 ml of PBS, then added 250 ml solutions of the

Optical microscopy
The confocal fluorescence, interference reflection, and differential interference contrast (DIC) imaging were carried out with an Olympus FV1000 confocal microscope (Olympus) and the FV10-ASW 1.7 software (Olympus). The images were collected with an oil-immersion UPlanSApo 1.35 NA 606 objective at room temperature, except that the time-lapse tracking of live cells was carried out at 37uC. Acquisition parameters including laser transmissivity, pixel dwell time, pinhole size, photomultiplier tube gains, and image size were adjusted for optimal detection sensitivity and minimal photobleaching.
Confocal fluorescence imaging. EGFP and Alex-488 were excited using the 488 nm laser line of an Ar ion laser. DiI was excited using the 543 laser line of a He-Ne laser. Cy5 was excited using the 633 nm laser line. The emissions of all fluorophores within 10 nm of bandwidth around their emission maxima were collected, except that the emissions of Cy5 was collected with BS 20/80 filter for the emissions greater than 640 nm. For multichannel imaging, the excitation and emission collection were carried out sequentially for each channel.
Interference reflection imaging. The 458 nm laser line of an Ar ion laser (1-2% of transmissivity) was used as incident light. The reflections at the bandwidth of 440-470 nm were collected. The imaging focus was set to the vicinity of the interface between cells or ventral membranes and the glass surface. This interface was found by changing the focus until the maximal reflection was obtained.
DIC imaging. Either 488 nm or 543 nm laser line was used as incident light and the transmitted light was collected.
Time-lapse imaging. The confocal microscope was equipped with a temperature and CO 2 control chamber (Life Imaging Services, Switzerland) that maintained the entire microscope at 37uC, and saturated the sample with humidified air containing 5% CO 2 . The time-lapse imaging acquisition was automated with the FV10-ASW 1.7 software. Images were taken at the frame rate as indicated in figures and in captions of corresponding videos. Image series were converted into videos using ImageJ (Wayne Rasband, National Institutes of Health, http://rsb.info.nih.gov/ij/) and QuickTime Player 7 (Apple Inc.). indicates the separation between the filopodia adhesion in the cell lamellum and the distal section of the filopodium, which might be fracture that could have occurred during the sample preparation. C. The cell periphery of an exposed ventral cell membrane from a REF52-b3-integrin-EGFP cell (30 min after plating on FN coated glass) showed a b3-integrin cluster at the filopodium (pink arrows), circumferential stress fibers (cyan arrowheads) anchored to the substrate via the b3integrin clusters formed at the side of this filopodia adhesion, and the sidewise widening of the filopodia b3-integrin cluster (pink arrowheads) following the connected thick circumferential stress fibers in the cell lamellum. Scale bar: 5 mm (A-b & c, B, C), 10 mm (A-a). doi:10.1371/journal.pone.0107097.g007

Image Analysis
All images in 16 bit tiff format were processed with ImageJ or the FV10-ASW 1.6 Viewer (Olympus). When needed, images in original grey scale were contrast adjusted and applied false colours for presentation. Overlays of selected frames from the time-lapse sequences were performed in ImageJ. Overlays of images of different optical signals in the same field of view were performed with FV10-ASW 1.6 Viewer or ImageJ. Montages of the timelapse sequences were made in ImageJ.
Kymographs. The kymograph is essentially the time-lapse montage of the cell edge region of defined width, which shows the trace of the cell edge position over time. The protrusion or retraction of the lamellipodium corresponds to the continuous increase or decrease of the cell edge positions respectively in the kymograph (Fig. S2C). Thus, the kinetic values of the protrusion or retraction of the lamellipodium, as well as of the net cell edge advancement, can be calculated accordingly. Kymographs were produced using ImageJ by taking 3-pixel-wide line in the timelapse sequence. This 3-pixel-wide line was drawn perpendicular to the cell edge and to the direction of the cell edge advancement. The duration (t) and distance (d) in the movement of the lamellipodium were determined in the kymograph along the slopes as the x (time) and y (distance) values respectively, and the speed of movement was calculated accordingly as d/t (Fig. S2C). The quantification of the kinetics of lamellipodia used the kymographs based on DiI fluorescence or DIC signals. For the quantification of the net advancement speed of the cell edge, DIC, DiI fluorescence and diffusive intracellular EGFP background fluorescence were used alone or in combinations for an accurate identification of the cell edge position.
Quantifications of cell-substrate adhesions. For the cellsubstrate adhesions identified as EGFP-b3-integrin clusters or vinculin immunostains, the regions of interest were cropped from the original images. The signal level of the background (non adhesion areas) in a cropped image was measured with the intensity profile along a line intersecting the adhesion. The original cropped images were then thresholded to only retain the adhesion signals. Alternatively for the images with fluorescence signals of the cell-substrate adhesions, the thresholding of the cropped images was done directly via the ''make binary'' routine in ImageJ. The adhesions in the threshold images were analyzed as particles (''analyze particles'' routine in ImageJ) and we quantified their areas, widths, lengths and aspect ratios. The area was measured based on the total pixels of an adhesion. The length and width of an adhesion were determined respectively as the distances along the primary and secondary axis of the best fit ellipse. The aspect ratio was calculated as the length/width of the fitted ellipse.
Quantification of the fluorescence intensity of the cellsubstrate adhesions. To quantify the fluorescence intensity, the b3-integrin-EGFP or vinculin clusters in the original image (in 8 bit format) were outlined by restoring the corresponding particle outlines in the thresholded image. After choosing ''mean gray value'' in the ''set measurements'' function, the ''analyze particles'' routine in ImageJ was used to measure the average fluorescence intensity, which is the mean grey value within the selection (the sum of the gray values of all the pixels in the selection divided by the number of pixels). The fluorescence intensity profile of the b3integrin-EGFP cluster was generated using ImageJ with a 3-pixelwide line drawn radially along the length of the filopodia adhesion in the last frame of its tracked sequence. The spatial start and end points of the measurement for the intensity profile correspond to the distal and proximal ends of the filopodia integrin cluster respectively.
Identification of filopodia adhesions in ventral cell membrane samples. They were determined as substrate adhesions in association with filopodia, or within ,10 mm behind the filopodia containing cell edge (Fig. 7). Focal adhesions are characteristically associated with the radial oriented long dorsal stress fibers ( [73] and Fig. S5C), and that the large focal adhesions in the early stage of spreading fibroblasts were located closer to the center of the cell (.10 mm from the cell edge).

Statistics
Data analysis was carried out in Excel (Microsoft). Significance was determined using the t-test for the independent data groups of unequal variance. The P values (two tail) were given in the text. The P values less than 0.05 were considered significant. Figure S1 Growth kinetics of filopodia base adhesions. A. Selected frames from the confocal microscopy time-lapse sequence of a growing b3-integrin-EGFP cluster associated with the filopodium base in a b3-integrin-EGFP (green) expressing REF52 fibroblast on FN coated glass (5 min-25 min after plating, Video S2). The cell edge and filopodium were visualized by the differential interference contrast (DIC) signal (grey). Formed behind the lamellipodium, the distal end of the filopodia base adhesion was in contact with the lamellipodium from the beginning of its formation. Along with the advancement of the lamellipodium, the filopodia base adhesion grew towards the filopodium tip. The dashed rectangle (top of the right cell image) was magnified in the left column (a-f). The white dashed vertical line is the reference line indicating the frontal elongation of the filopodia b3-integrin-EGFP cluster towards the filopodium tip by tightly following the advancing lamellipodium (a, b). This filopodia b3-integrin-EGFP cluster continuously widened (c, d, e, f) during its long duration of contact with the lamellipodium. Black arrows indicate the mobile non-adherent distal section of the filopodium, which was bended and recycled after the lamellipodium paused at the distal tip of the filopodia adhesion without net advancement. White arrows point to the newly formed b3-integrin-EGFP clusters extending circumferentially at the sides of the filopodia adhesion. White curves indicate the manually drawn contour of the lamellipodium based on the DIC contrast. The colored frames in A correspond to the colored phases as classified in B, i.e. the time period in which the b3-integrin-EGFP cluster contacts the advancing lamellipodium (yellow) and the events during pausing of the lamellipodium (pink) respectively. On the right column, the b3-integrin-EGFP images from different time points are overlaid (cyan-early, white-later). B and C. Quantified changes of size (B) and average fluorescence intensity (C) over time in the filopodia b3-integrin-EGFP cluster in A. The quantitative analyses were carried out and presented in the same way as in Fig. 1B 1) In lamellipodia associated with filopodia in neuronal growth cones [63]. (2) In isotropic spreading mouse embryonic fibroblast without filopodia [37].  Fig. 4D-c, Video S6) in relation to the lamellipodium on soft (7.4 kPa) PAA gel. The analyses were performed and presented in the same way as for the filopodia adhesion in Fig. 1. D. The spatial distribution of b3integrin in this filopodia adhesion. As described in Fig. 2E, the fluorescence intensity profile was generated along the long axis of the filopodia integrin cluster (insert: last frame of the tracking sequence, 3 mm67 mm).  Figure S5 The growth kinetics of the adhesion formed at the lamellipodium type of cell edge (A and B). A. Selected frames from the confocal microscopy time-lapse sequence of a growing b3-integrin-EGFP cluster in a b3-integrin-EGFP (green) expressing REF52 cell on FN coated glass (7 min-22 min after plating). The lamellipodium cell edge was identified by the DIC signal (grey). Formed in the lamellipodium, the adhesion (white arrows in b) grew into the cell lamellum. The dashed square in a was magnified in b at selected time points. B. Quantified growth of the adhesion initiated in the lamellipodium in its width (blue), length (red), area (black) and aspect ratio (grey). The same scales in vertical axes were used as for the filopodia adhesion in Fig. 1B for comparison. The growth characteristics of adhesions of lamellipodial origin were different from those of filopodia adhesions during the tracked early growth (,900 s). The length of the adhesion initiated at the lamellipodium type of cell edge was small and grew slowly without a stage of rapid increase, so that its aspect ratio only fluctuated slightly at very low values. C. The cytoskeletal organization at adhesions originated from lamellipodia. a. View of the exposed ventral surface of the cell edge region with no filopodia. This HFF cell (40 minutes plated on FN coated glass) partially retained its dorsal cell membrane structures (white asterisk). b,c,d. Magnified views in respective signals of the dashed rectangle region in a. The dorsal location of the remnant dorsal cell membrane structures (b and c, white asterisks) was confirmed by the higher membrane and actin fluorescences and the bright interference reflection signal in that region. The remnant dorsal cell membrane structures were tethered to the proximal sections of multiple radial oriented dorsal stress fibers. These dorsal stress fibers were anchored at their distal ends (d, pink arrows) in adhesions (c, pink arrows) of the size of focal complexes, which were originated from the lamellipodia type of cell edge. Scale bars: 1 mm (A-b), 2 mm (C-b), 5 mm (A-a, C-a).

(TIF)
Video S1 The development of the filopodia tip adhesion in a spreading REF52-b3-integrin-EGFP cell. (Fig. 1). Overlays of membrane lipid fluorescence (red) and b3-integrin-EGFP (green). The white rectangle region corresponds to the magnified view in Fig. 1A left column. 44 min 20 s-61 min 20 s after plating the cell on FN coated glass. The time-lapse seqeunce was collected at 20 s/frame with an Olympus FV1000 confocal microscope (Olympus). The video is shown at 5 frame/s.

(MOV)
Video S2 The development of the filopodia base adhesion in a spreading REF52-b3-integrin-EGFP cell. (Fig.  S1). Overlays of DIC signal (grey) and b3-integrin-EGFP fluorescence (green). The white rectangle region corresponds to the magnified view in Fig. S1A left column. 5 min-25 min after plating the cell on FN coated glass. The time-lapse seqeunce was collected at 20 s/frame with an Olympus FV1000 confocal microscope. The video is shown at 5 frames/s.

(MOV)
Video S3 The lamellipodium exhibited cyclic protrusions and retractions in the proximity of the growing filopodia adhesion in a spreading REF52-b3-integrin-EGFP cell. (Fig. 3). Overlays of membrane lipid fluorescence (red) and b3-integrin-EGFP (green). The white rectangle region corresponds to the filopodia adhesion region analyzed in Fig. 3A-C. 20 min-47 min after plating the cell on FN coated glass. The time-lapse seqeunce was collected at 18 s/frame with an Olympus FV1000 confocal microscope. The video is shown at 5 frames/s.

(MOV)
Video S4 The abolishment of the cyclic protrusions and retractions of the lamellipodium in a spreading REF52-b3-integrin-EGFP cell by MLCK inhibition with 30 mM of ML-7. (Fig. 4A). Membrane lipid fluorescence signal (red). 25 min 3 s-29 min 58 s after plating the cell on FN coated glass. The white rectangle region corresponds to the cell edge region represented in montage. The time-lapse seqeunce was collected at 5 s/frame with an Olympus FV1000 confocal microscope. The video is shown at 5 frames/s. (MOV) Video S5 Significant ruffling of the lamellipodium in a spreading REF52-b3-integrin-EGFP cell on soft PAA substrate due to the lack of effective surface adherence. Video S6 The restored development of filopodia adhesion in the spreading REF52-b3-integrin-EGFP cell on soft PAA substrate with the locally suppressed ruffling of the lamellipodium. (Fig. 4D). Overlays of membrane lipid fluorescence (red) and b3-integrin-EGFP (green). The white rectangle region corresponds to the filopodia adhesion region analyzed. 121 min 26 s-136 min 26 s after plating the cell on FN coated PAA substrate. The time-lapse seqeunce was collected at 30 s/frame with an Olympus FV1000 confocal microscope. The video is shown at 5 frames/s.

(MOV)
Video S7 The rapidly disassembly of a filopodia adhesion when having very brief contact with the advancing lamellipodium in a spreading REF52-b3integrin-EGFP cell. (Fig. 5A). Overlays of membrane lipid fluorescence (red) and b3-integrin-EGFP (green). The white rectangle region corresponds to the filopodia adhesion region analyzed. 40 min 20 s-53 min 40 s after plating of the cell on FN coated glass. The time-lapse seqeunce was collected at 20 s/frame with an Olympus FV1000 confocal microscope. The video is shown at 5 frame/s. Frame size: 1.3 mm64.4 mm.

(MOV)
Video S8 During the fast spreading of a REF52-b3integrin-EGFP cell, the fast advancing cell edge could rapidly surpass the filopodia adhesion and leave it behind in the cell lamellum, resulting in its rapid disassembly (at the horizontal arrow, Fig. 6B-b, I). (Fig. 6B). Alternatively, the fast advancing cell edge could abruptly pause for long duration at the filopodia adhesion, resulting in its stable maturation (at the vertical arrow, Fig. 6B-b, II). Overlays of membrane lipid fluorescence (red) and b3-integrin-EGFP (green). 5 min-9 min 36 s after plating the cell on FN coated glass. The time-lapse seqeunce was collected at 12 s/frame with an Olympus FV1000 confocal microscope. The video is shown at 5 frames/s. (MOV) Video S9 The rapid disassembly of filopodia adhesions due to their brief contact time with the fast advancing lamellipodium in a fast spreading HFF cell. (Fig. S4C). Interference reflection signal. 8 min-28 min after plating the cell on FN coated glass. Red arrows: the dissembling filopodia adhesions. The time-lapse seqeunce was collected at 30 s/frame with an Olympus FV1000 confocal microscope. The video is shown at 5 frames/s.

(MOV)
Video S10 The fast advancing cell edge paused at the distal tip of the filopodia adhesion in a spreading HFF cell. (Fig. S4D). Interference reflection signal. The white square region corresponds to the filopodia adhesion region shown in Fig.  S4D-b. Red arrow: the filopodia adhesion. 41 min-50 min after plating the cell on FN coated glass. The time-lapse seqeunce was collected at 20 s/frame with an Olympus FV1000 confocal microscope. The video is shown at 5 frame/s. (MOV)