Calcium Rises Locally Trigger Focal Adhesion Disassembly and Enhance Residency of Focal Adhesion Kinase at Focal Adhesions*

Focal adhesion kinase (FAK) activity and Ca2+ signaling led to a turnover of focal adhesions (FAs) required for cell spreading and migration. We used yellow Cameleon-2 (Ycam), a fluorescent protein-based Ca2+ sensor fused to FAK or to a FAK-related non-kinase domain, to measure simultaneously local Ca2+ variations at FA sites and FA dynamics. Discrete subcellular Ca2+ oscillators initiate both propagating and abortive Ca2+ waves in migrating U87 astrocytoma cells. Ca2+-dependent FA disassembly occurs when the Ca2+ wave reaches individual FAs, indicating that local but not global Ca2+ increases trigger FA disassembly. An unexpectedly rapid flux of FAK between cytosolic and FA compartments was revealed by fluorescence recovery after photobleaching studies. The FAK-Ycam recovery half-time (17 s) at FAs was slowed (to 29 s) by Ca2+ elevation. FAK-related non-kinase domain-Ycam had a faster, Ca2+-insensitive recovery half-time (11 s), which is consistent with the effect of Ca2+ on FAK-Ycam dynamics not being due to a general modification of the dynamics of FA components. Because FAK association at FAs was prolonged by Ca2+ and FAK autophosphorylation was correlated to intracellular Ca2+ levels, we propose that local Ca2+ elevations increase the residency of FAK at FAs, possibly by means of tyrosine phosphorylation of FAK, thereby leading to increased activation of its effectors involved in FA disassembly.

integrins span the plasma membrane and mediate interactions between extracellular matrix components and cytoskeletal proteins. In migrating fibroblasts, FAs are mainly immobile, serving as traction points, except in restricted retractile cell edges, where they display centripetal movements (1). This local regulation of FA movement involves asymmetric signal transduction and probably directs cell polarization and migration. Despite the observation of an increasing Ca 2ϩ gradient from front to rear in migrating fibroblasts (2,3), Ca 2ϩ is not often suggested to be a polarity signal in chemotaxis. This is surprising, given the large number of potential Ca 2ϩ -sensitive targets involved in integrin signaling and cell migration (4 -8) and the tight spatio-temporal regulation of Ca 2ϩ elevations (9). In support of a spatially restricted action of Ca 2ϩ during cell migration, Ca 2ϩ oscillations induce cell rear retraction in lymphocytes and keratinocytes (10,11). Using fluorescent protein fusion constructs targeted to FAs, we showed recently (12) that Ca 2ϩ elevations trigger local disassembly of FAs in human U87 astrocytoma cells. Here, we again made chimeric proteins using focal adhesion kinase (FAK) as an FA-targeting module and yellow Cameleon-2 (Ycam), a fluorescent protein-based Ca 2ϩ sensor (13) to measure both local Ca 2ϩ variations at FAs by fluorescence resonance energy transfer (FRET) between enhanced cyan fluorescent protein (ECFP) and enhanced yellow fluorescent protein (EYFP), and FA dynamics. Here, we describe temporal and spatial correlations linking local Ca 2ϩ elevations to disassembly of individual FAs.
In the hierarchical organization of FA-associated proteins, FAK appears as an early component in FA formation (14). The integrin/adhesion-dependent increase in FAK Tyr 397 autophosphorylation relies on intermolecular FAK transphosphorylation (15) mediated by the formation of integrin clusters. The central role of FAK in the formation of a functional FA is also emphasized by its scaffolding function, allowing direct interaction and translocation of signaling proteins such as Src, growth factor receptor binding protein-2, paxillin, phospholipase C-␥, phosphatidylinositol 3-kinase, and p130cas (16) toward FAs. Furthermore, FAK interacts with FA structural proteins, including integrins, talin, ␣-actinin and tensin, and via paxillin with both vinculin and F-actin (17). In FAK-deficient cells, reduced motility is accompanied by an increased number of FAs (18), suggesting that FAK tyrosine kinase activity is involved in the regulation of FA turnover (19). Expression of the dominant negative FAK-related non-kinase domain (FRNK) inhibits integrin-stimulated migration and phosphorylation of endogenous FAK and other FA components such as paxillin and tensin (20 -22). Thus, FAK activity likely governs FA dynamics and, hence, motility. However, no studies have yet addressed the possible regulation of FAK signaling by modification of its partitioning between FAs and the cytosol.
These considerations led us to examine FAK-Ycam and FRNK-Ycam molecular dynamics using fluorescence recovery after photobleaching (FRAP). We report a rapid exchange of both FAK-Ycam and FRNK-Ycam between cytosolic and FA compartments. FRNK-Ycam, which induced a decrease in tyrosine phosphorylation content at FA, had a shorter recovery half-time compared with that of FAK-Ycam. The local Ca 2ϩ increases triggering FA disassembly prompted us to investigate the effect of Ca 2ϩ on the dynamics of FAK-Ycam and FRNK-Ycam. Ca 2ϩ elevation slowed the recovery half-time of both FAK-Ycam and FAK-GFP at FAs while having no effect on FRNK-Ycam. Several studies report that FAK autophosphorylation levels are increased by Ca 2ϩ elevation (23)(24)(25); in agreement, we found that rises in Ca 2ϩ and Ca 2ϩ buffering, respectively, increased and decreased FAK Tyr 397 phosphorylation at FAs. Thus, the extended residency of FAK at FAs when Ca 2ϩ is elevated correlates with increased FAK tyrosine phosphorylation, FA disassembly, and cell-edge retraction. Therefore, it is possible that enhanced residency of FAK at FAs triggered by Ca 2ϩ rises might play a role in Ca 2ϩ -dependent FA disassembly. Taken together, our results suggest a tight regulation of FAK association at FAs, which likely directs the activation status of FAK-associated targets involved in FA disassembly.

EXPERIMENTAL PROCEDURES
Reagents and Cells-Eagle's minimal essential medium, fetal calf serum (FCS), HEPES, L-glutamine, penicillin, streptomycin, gentamycin, and trypsin-EDTA solution were obtained from Invitrogen. Fura Red-AM, BAPTA-AM, and pluronic acid F-127 were obtained from Molecular Probes. Matrigel and the monoclonal antibody (mAb) directed against the kinase domain of FAK were obtained from BD Biosciences. The human U87 astrocytoma cell line was obtained from American Type Cell Culture. Cells were maintained at 37°C in a humidified incubator gassed with 5% CO 2 in air on type I collagen-coated (0.06 mg/ml) plastic dishes in Eagle's minimal essential medium supplemented with 10% heat-inactivated FCS, 0.6 mg/ml glutamine, 200 units/ml penicillin, 200 units/ml streptomycin, and 0.1 mg/ml gentamycin.
Expression Plasmids and Transfection-As described previously (12), fluorescent FA-targeted protein was made by fusion of FAK cDNA from human T lymphocytes (pCDM8-FAK plasmid (26), kindly provided by S. B. Kanner) next to the 3Ј end of Ycam (pcDNA3-Ycam2 plasmid (13), kindly provided by R. Y. Tsien). FRNK-Ycam was made by following a similar protocol. All constructs were verified by sequencing. A FAK-GFP plasmid (34) was kindly provided by L. H. Romer. The plasmids were isolated (JetStar Plasmid kit, Genomed) before transfection by electroporation. Cells were selected 24 h later using 800 g/ml G418 (Sigma) and maintained with 400 g/ml G418. To obtain Ͼ80% expressing cells, cells were sorted using a FACStar cell sorter (BD Biosciences) before use.
Measurement of Intracellular Ca 2ϩ and Focal Adhesion Dynamics-Global intracellular Ca 2ϩ levels were measured using Fura Red by confocal microscopy (Kr/Ar laser, Bio-Rad 1024) on an inverted microscope (Nikon Eclipse TE300), with 40ϫ oil-immersion CFI Plan-Fluor objective (numerical objective 1.3) simultaneously with FA dynamics using emitted fluorescence of EYFP (12). Cells (5 ϫ 10 4 cells/ml) expressing FAK-Ycam or FRNK-Ycam were grown to sub-confluence on Matrigel in Petri dishes in which a 2-cm diameter hole had been cut in the base and replaced by a thin (0.07 mm) coverslip. After 48 or 72 h, cells in the culture medium were loaded with 10 M Fura Red-AM and 0.03% pluronic acid (45 min) in the incubator and then washed 2ϫ with Ringer's solution (140 mM NaCl, 5 mM KCl, 2 mM CaCl 2 , 2 mM MgCl 2 , 10 mM HEPES, 11 mM glucose, pH 7.4). Imaging of single U87 cells in Ringer's solution containing 10% FCS was done at 32°C (488 nm excitation), and emitted fluorescence was collected simultaneously at 585 Ϯ 10 nm for Ca 2ϩ (Fura Red) and at 522 Ϯ 16 nm for FA dynamics (EYFP) at 10-s intervals, usually for 30 min. Images were taken at the bottom of the cell surface. FA dynamics were analyzed using NIH Image, with fluorescence intensities from manually outlined, FA-containing regions of interest of ϳ3-5 m 2 , as described previously (12). FA movement is illustrated in the figures using a time-lapse, three-color overlay representation (1); images are shown after color inversion. Intracellular Ca 2ϩ measurements based on FRET between ECFP and EYFP and restricted to FAs were made using FAK-Ycam on a confocal system as above, except that excitation was at 440 from a He/Cd laser. Simultaneous measurements of local Ca 2ϩ measurements (FRET between ECFP and EYFP) and FA dynamics (EYFP emission) were done at the European Molecular Biology Laboratory on a temperature-controlled stage (32°C) of a Zeiss LSM 510 confocal system equipped with a 60ϫ oil-immersion objective (1.4 numerical objective); excitation was at 413 nm from a Kr laser. Cells expressing FAK-Ycam were grown to sub-confluence on Matrigel in Petri dishes when using the Bio-Rad 1024 or in Labtech chambers when using the Zeiss LSM 510. Single cells were imaged in Ringer's solution plus 10% FCS, and emitted fluorescence was collected simultaneously at 485 nm for ECFP and at 530 nm for EYFP at 0.5-1 Hz from ϳ20 -25 m 2 regions of interest containing individual FAs. Thus, Ycam FRET-dependent Ca 2ϩ measurements are based upon FAK-Ycam expressed both in FAs (which have a smaller surface, typically 3-5 m 2 ) and in the cytosol surrounding FAs. Intracellular Ca 2ϩ variations were represented by the ratio of EYFP/ECFP emission (R530/485) and normalized to fluorescence intensity in the regions of interest at the beginning of the recording. Images were analyzed as above.
Fluorescence Recovery after Photobleaching Experiment-As described previously (27), FRAP experiments were done on a custom confocal microscope at the European Molecular Biology Laboratory at 32°C with simultaneous excitation at 488 nm and 514 nm (100 mW Ar ion laser). For quantitative measurements, the photobleached region corresponded to a rectangle enclosing the selected FA or cytoplasmic region. Fluorescence within the rectangle was measured at low laser power before bleach and then photobleached with full laser power (100% power, 100% transmission) for ϳ6 s (which effectively reduced the fluorescence to background levels in fixed cells; data not shown). Recovery was followed with low laser power at ϳ1-s intervals until a steady plateau was reached. Negligible bleaching occurred during imaging of recovery, as verified in control experiments. Time was corrected by setting time zero as equal to the half-time of the bleaching period. The post-bleach intensities were normalized slightly upward to correct for the loss of total fluorescence because of photobleach (typically Ͻ2%). Fluorescence recovery time courses were normalized to pre-bleach intensity. We compared relative recovery rates for FAK-Ycam, FAK-GFP, and FRNK-Ycam in various conditions, using the half-time for recovery of fluorescence toward the asymptote. Mobile and immobile fractions were calculated by comparing the intensity ratio in the bleached area just before the bleach and after recovery. FRAP data are given as mean Ϯ S.D. for nFAs.
Immunostaining-FAK-Ycam and FRNK-Ycam cells were rinsed with phosphate-buffered saline (PBS) and fixed 15 min with 3% paraformaldehyde at room temperature. After three washes with PBS, cells were treated for 10 min with 0.2% Triton X-100 in PBS, 0.2% bovine serum albumin (BSA) and incubated 30 min with PBS/3% BSA. Cells were washed three times with PBS, 0.2% BSA and incubated 1 h with either the anti-FAK P-Tyr 397 mAb (1/1000) or the anti-P-Tyr Ab (1/1000) in PBS, 0.2% BSA. For FAK Tyr 397 or total Tyr staining, cells were washed three times and incubated 1 h with a TRITC-labeled goat anti-mouse Ab (1/100). Fluorescence was observed with a Bio-Rad confocal system and a 40ϫ objective, as above. EYFP and TRITC were excited at 488 and 568 nm, respectively, and fluorescence was collected at 522 and 585 nm, respectively.

Local Active Ca 2ϩ Oscillators Trigger Propagating Ca 2ϩ
Waves-Recently, we showed that Ca 2ϩ spikes observed in migrating U87 cells represent a Ca 2ϩ wave that spreads throughout the cell (28). Here, Ca 2ϩ measurements restricted to FAs were made using FAK-Ycam. U87 cells having stable expression of FAK-Ycam were grown to sub-confluence on Matrigel and imaged in Ringer's solution containing 10% FCS, conditions necessary to observe Ca 2ϩ spikes (12) and functional integrins. We observed clear kinetic differences in Ca 2ϩ elevations at discrete FAs (Fig. 1A), which is in agreement with a propagated wave. The delays measured for the appearance of Ca 2ϩ spikes at individual FAs located in different parts of a cell ( Fig. 1C) allowed calculation of a Ca 2ϩ wave speed of 20 Ϯ 10 m/s (n ϭ 48 waves, 7 cells). An interesting feature is the presence of active Ca 2ϩ oscillators, which reliably trigger Ca 2ϩ wave propagation. These sites were defined as the nearby FAs displaying the earliest increase in Ca 2ϩ during spiking (Fig. 1, compare red and light-blue FAs) and were characterized by a distinct and relatively constant frequency of oscillation. Between one and three active Ca 2ϩ oscillators were detected in different cells (n ϭ 22). Both regenerative, propagating Ca 2ϩ waves (which couple to remote conveying Ca 2ϩ oscillators; see Fig. 1B, first four spikes) and subcellularly restricted or abortive Ca 2ϩ waves (which do not propagate throughout the cell; see Fig. 1B, spikes 5-7 of red FA) appear to be generated by active oscillators. Both types of waves probably share common underlying mechanisms, because the propagation speed for abortive Ca 2ϩ waves (21 Ϯ 12 m/s; n ϭ 52 waves, 9 cells) was the same as for regenerative waves. The location of active Ca 2ϩ oscillators seems not to be correlated with the direction of cell migration, as we have observed polarized cells having active Ca 2ϩ oscillators at both poles (not shown). However, it should be noted that our experiments were done without any directed signals (low density plating, homogeneous serum stimulation). Note also that our methods did not allow us to resolve whether FAs are the sites where integrin-dependent Ca 2ϩ signaling is first generated. Nevertheless, we used FAK-Ycam, which concentrates FAs, to measure local Ca 2ϩ rises around individual FAs in an effort to demonstrate the compartmentalized nature of Ca 2ϩ elevations.
Local Subcellular Ca 2ϩ Increases Trigger Disassembly of Neighboring FAs-We also reported that Ca 2ϩ spikes trigger FA disassembly (12), based upon simultaneous measurements of global Ca 2ϩ variations and dynamics of FAs or actin stress fibers. We tested whether the spatially restricted Ca 2ϩ increases described above were capable of triggering adjacent FA disassembly (Fig. 2). Color overlays (blue, red, green) of three sequential images of EYFP emission ( Fig Fig. 2D, red trace). Although this early Ca 2ϩ increase appears to be weakly propagated to region 1, disassembly of FA 1 occurs later at 370 s, coincident with a strong local Ca 2ϩ increase in region 1 (Fig. 2B, green trace; Fig. 2C, red trace) which hardly propagated to region 2. These data demonstrate that disassembly of individual FAs is associated with local increases in Ca 2ϩ . This spatial and temporal correlation was observed in all cells where Ca 2ϩ increases in different parts of the cell were clearly separated in time, with the latency between the onset of local Ca 2ϩ increases and FA disassembly being always shorter (15 Ϯ 2 s, n ϭ 7 FAs, 6 cells) when compared with global Ca 2ϩ increases measured from the entire cell (39 Ϯ 8 s, n ϭ 7 FAs, 6 cells).
We reported previously (28) that integrin-dependent Ca 2ϩ oscillations in U87 cells were reduced by depletion of Ca 2ϩ stores and abolished by non-selective inhibition of Ca 2ϩ channels by Cd 2ϩ and La 3ϩ . However, Ga 3ϩ , which inhibits stretch-dependent Ca 2ϩ channels, did not affect the generation of Ca 2ϩ oscillations in U87 cells (not shown), unlike in keratinocytes (11).
To show directly that Ca 2ϩ increases because of Ca 2ϩ entry cause FA disassembly, we bathed FAK-Ycam cells in Ca 2ϩ -free external medium (with 10% FCS) containing 5 mM EGTA (to empty internal Ca 2ϩ stores) and 5 M ionomycin (Fig. 3). Upon local and repeated microperfusion of the same solution (but containing, in addition, 2 mM Ca 2ϩ ), transient increases in Ca 2ϩ were observed (using Fura Red; see Fig. 3D, solid trace) to trigger FA disassembly (Fig. 3D, dotted traces). Note the retraction of the cell edge after Ca 2ϩ application (Fig. 3B). Although we cannot rule out that the experimental protocol used might give rise to Ca 2ϩ -sensitive outside-in integrin signaling, on balance, given the present knowledge concerning how external Ca 2ϩ affects integrin function (for review, see Ref. 70), we feel it is quite plausible that the observed elevations in Ca 2ϩ provoked by ionomycin-assisted Ca 2ϩ entry trigger FA disas- sembly in a qualitatively similar way to integrin-dependent Ca 2ϩ oscillations (Fig. 2).
FAK Molecular Dynamics Are Characterized by a Fast Equilibrium between FAs and Cytosol-In motile U87 cells, we also observed predominantly static FAs, as found in fibroblasts (11), together with restricted zones where FAs disassembled and formed, respectively, at the cell rear and at the migration front (12). FAs need to be relatively stable structures in time and space to support traction of the cytoskeleton necessary to propel the cell body (11). Cell-edge retraction triggered by FA disassembly stops when the next stable FA is encountered. We occasionally observed a slow redistribution of FAK-Ycam fluo-rescence from a disassembling FA to an adjacent FA (Fig. 4). As before, color overlays (blue, red, green) of three sequential images of EYFP emission (Fig. 4A) distinguish immobile FAs (black) from motile FAs (rainbow). This exchange of FAK (Fig.  4A, right panels), which occurs over several minutes, is seen as a loss of FAK-Ycam EYFP fluorescence from a disassembling FA 1 (Fig. 4A, dotted circle in upper right panel; Fig. 4B, green line) and an increase of EYFP fluorescence in an adjacent, (initially) stable FA 2 (Fig. 4A, upper right panel, solid circle; Fig. 4B, blue trace).
To test whether the disappearance of FAK-Ycam fluorescence during FA disassembly corresponds to the dissociation of a stable complex containing FAK and FA structural components, as suggested by Fig. 4, the molecular dynamics of FAK-Ycam exchange between the cytosolic and FA compartments were studied by FRAP (Fig. 5). Experiments were made to determine the turnover of FAK-Ycam at immobile FAs. The EYFP moiety of Ycam was bleached at individual FAs, and subsequent time-lapse imaging of bleached regions allowed visualization of FAK-Ycam movement during recovery (Fig. 5B,  second column). The recovery of FAK-Ycam at FAs was surprisingly fast, with a half-time of 17 Ϯ 4 s (Fig. 5D, n ϭ 12 FAs), which, together with the 79 Ϯ 4% recovery, indicates the presence of a small immobile fraction. This supports the idea that there is a large and rapid exchange flux of FAK between cytosolic and FA compartments, with FAK molecular dynamics being much faster compared with both FA formation and disassembly (which take Ͼ5 min). FRAP experiments done on small areas of the peripheral cytosol revealed a FAK-Ycam recovery half-time of 3 Ϯ 1 s (Fig. 5C, n ϭ 13), which is much faster than at FAs and is most likely diffusion-limited. This result indicates that additional factors, for example binding to various FA components, are involved in regulating FAK exchange between FAs and the cytosol. Therefore, the fast exchange of FAK between cytosolic and FA compartments contrasts with the stability of FAs and suggests that FAK forms transient interactions with FA components.  terminal domain, the kinase domain, and the Tyr 397 autophosphorylation site, but including the C-terminal FA-targeting sequence (20). Accordingly, both the total amount of tyrosine phosphorylation (Fig. 6A) and FAK Tyr 397 phosphorylation (Fig. 6B) were decreased in FRNK-Ycam-transfected cells compared with adjacent non-transfected cells. To test whether FAK and FRNK displayed the same molecular dynamics, FRAP experiments were done using FRNK-transfected cells. The recovery of bleached FRNK-Ycam at FAs (Fig. 5B, first row) was significantly faster than for FAK-Ycam (Fig. 5B, second row), with a recovery half-time of 11 Ϯ 3 s (Fig. 5D, n ϭ 13 FAs), whereas the immobile fraction was similar (16 Ϯ 6%). However, no significant differences were noted between FAK-Ycam and FRNK-Ycam dynamics after cytosolic bleaching, with a FRNK recovery half-time of 2 Ϯ 1 s (Fig. 5C, n ϭ 8). These data indicate that the faster recovery of FRNK at FAs is not due to differences in cytosolic diffusion of FAK compared with FRNK, and that the slower recovery of FAK is due to a longer residency time at FAs.
Calcium Elevation Prolongs FAK Association at FAs-Only a few studies have reported a correlation between Ca 2ϩ elevations and FAK-mediated increases in tyrosine phosphorylation (23)(24)(25). In U87 cells, thapsigargin-induced Ca 2ϩ increases produce a rapid (Ͻ30 s) enhancement of FAK Tyr 397 phosphorylation (12), as also seen for ionomycin-induced Ca 2ϩ elevation (not shown). In agreement, Ca 2ϩ buffering with BAPTA induced a dramatic reduction of FAK Tyr 397 phosphorylation at FAs (Fig. 6C). Because enhanced migration associated with sustained FA turnover is dependent upon FAK activity (18,19,29), activation of FAK targets, and FA-localization of FAK (22,30), it is possible that increases in Ca 2ϩ , which trigger FA disassembly, might also have regulatory effects on the activation and/or localization of FAK.
Therefore, we used ionomycin to artificially trigger Ca 2ϩ elevation and then examined the molecular dynamics of FAK-Ycam and FRNK-Ycam in FA and cytosolic compartments by FRAP (Fig. 5). Despite extensive disassembly of FAs observed in most cells after ionomycin application (not shown), sufficient remaining stable FAs allowed us to carry out FRAP experiments. The FA residency time of FAK-Ycam was significantly increased (Fig. 5B, third row) after ionomycin treatment, with a recovery half-time of 29 Ϯ 7 s, compared with 17 Ϯ 4 s in control conditions (Fig. 5F, n ϭ 18 FAs). Furthermore, the immobile FAK-Ycam fraction was slightly but significantly increased, to 32 Ϯ 5%, compared with 21 Ϯ 4% in control conditions. By contrast, the exchange of FRNK-Ycam between the cytosol and FAs was not affected by an increase of Ca 2ϩ (recovery half-time ϭ 11 Ϯ 3 s, immobile ϭ fraction 17 Ϯ 9%; Fig.  5E, n ϭ 10 FAs). Likewise, in the presence of ionomycin, the cytosolic recovery half-time of FAK-Ycam (3 Ϯ 2 s; Fig. 5F, n ϭ 13) was the same as in control (Fig. 5C). Thus, increases in Ca 2ϩ slow FAK-Ycam dynamics at FAs, which are possibly linked to enhanced tyrosine phosphorylation, but not those of FRNK-Ycam, indicating that Ca 2ϩ elevation does not result in a nonspecific modification of the dynamics of components at FAs. Furthermore, the FAK-Ycam recovery half-times (tens of seconds) are in sharp contrast with the much slower kinetics (Ͼ5 min) of FA disassembly (and formation), suggesting that a long-lived dissociation (or association) of FAK from (with) an FA complex occurs during these processes.
To verify that the Ca 2ϩ -sensitivity of FAK-Ycam dynamics at FAs was unrelated to the conformational changes of Ycam induced by Ca 2ϩ binding (13) after ionomycin treatment, we repeated FRAP experiments on cells transfected with FAK-GFP (34). As illustrated, both FAK-GFP and endogenous FAK were revealed in Western blots, with localization of FAK-GFP in FAs (Fig. 7A). Compared with FAK-Ycam dynamics in untreated cells (recovery half-time ϭ 17 Ϯ 4 s, immobile fraction ϭ 21 Ϯ 4%; see Fig. 5D), FRAP experiments showed that FAK-GFP dynamics at FAs were faster with recovery halftimes of 10 Ϯ 3 s (Fig. 7B, n ϭ 34 FAs), with a similar immobile fraction of 24 Ϯ 7%. These faster dynamics may be related to the smaller size of FAK-GFP compared with FAK-Ycam. After treatment with ionomycin, FAK-GFP recovery half-times (Fig.  7B) were significantly increased (22 Ϯ 5 s; n ϭ 26 FAs), with a small increase also in the immobile fraction (34 Ϯ 6%). These slowing effects of increased intracellular Ca 2ϩ on FAK-GFP dynamics are similar to those found for FAK-Ycam after ionomycin treatment (recovery half-time ϭ 29 Ϯ 7 s, immobile fraction ϭ 32 Ϯ 5%; see Fig. 5).

DISCUSSION
Despite the relative immobility of most FAs and their life span on the order of at least several minutes, this study shows that the association/dissociation of FAK and FRNK with FAs is a fast dynamic process, with increases in Ca 2ϩ leading to longer association of FAK but not FRNK at FAs. This result is the first indication that FAK signaling might be regulated by modification of its partitioning between FAs and the cytosol. Note that both FA formation and disassembly have much slower kinetics, and that the molecular dynamics of FAK are about 10-fold faster compared with other FA proteins (31)(32)(33). Simultaneous measurements of local Ca 2ϩ variations at FAs and FA dynamics reveal that Ca 2ϩ -dependent FA disassembly occurred when the Ca 2ϩ wave, which is composed of transient Ca 2ϩ subcellu- FAK functions as a molecular adaptor interacting with both signaling and structural proteins (16), which is consistent with FAK being a multifunctional protein playing both signaling and structural roles (17). The continuous fast exchange of FAK-Ycam between cytosolic and FA compartments reported here emphasizes a signaling function for FAK, given the immobility of most FAs in migrating cells (1,19). In control conditions, after photobleaching FAK-Ycam, we found an unexpectedly fast recovery half-time of 17 s, which is the fastest so far described for an FA protein. This finding is consistent with recent work showing that FAK autophosphorylation is intermolecular, and that FAK cannot form stable homo-oligomers (15); rather, FAK requires external proteins for transient transphosphorylation and subsequent action on its targets. By comparison, recovery half-times after photobleaching of 2-5 min (31) or Ͼ10 min (32) have been reported for the integrin ␤ 3 subunit; for ␣-actinin, the recovery half-time after photobleaching was 2-3 min (32,33). Even if ␤ 3 integrins move slowly inwards at FAs (31), our results indicate that FAK and FRNK rapidly associate and dissociate from a more stable platform at FAs composed of at least integrins. In agreement with a predominant signaling role for FAK, chromophore-assisted laser inactivation of FAK at FAs does not lead to stress fiber detachment (34). Nevertheless, redundant interactions that characterize FA-associated proteins could underlie protein exchange without dissipation of FA architecture and might reconcile fast FAK dynamics with a structural function. Consistently, low affinity interactions between FA components and integrins have been described (35), which may facilitate protein-exchange dynamics. Note that chromophore-assisted laser inactivation of ␣-actinin, which exchanges between FAs and the cytosol more slowly than FAK, leads to disruption of integrincytoskeleton interactions (34). The fast dynamics of FAK-Ycam suggest that FAK exchange between FA and cytosolic compartments may be constant during the relatively slower process of FA disassembly (Ͼ5 min). Thus, FAK-Ycam fluorescence probably reflects the size of the integrin-containing platform with which FAK associates and which is more stable.
The identical ability of FAK and FRNK to localize at FAs, which are unaltered by the Ycam tag, is due to the FAT sequence (36) that contains binding sites for paxillin and talin (37,38), both of which seem to be involved in FA targeting. The shorter FA association time for FRNK compared with FAK may result from FRNK being unable to mimic FAK signaling or from a difference in affinity for FA components. Indeed, in FRNK, an integrin interaction site in the FAK N-terminal domain is absent (39), along with the kinase domain and the Tyr 397 autophosphorylation site responsible for recruitment and activation of Src-kinase (40).
In agreement with the predominant signaling hypothesis for FAK, the region between FAT and the kinase domain, as well as the N-terminal domain, are not necessary for adhesion-dependent FAK tyrosine phosphorylation and FAK-dependent paxillin tyrosine phosphorylation, and thus are probably not involved in FAK targeting to integrins and downstream signaling mediated by FAK (41). However, the FAT domain is necessary for FAK and FRNK to promote adhesion-dependent FAK tyrosine phosphorylation and for the dominant-negative function of FRNK (30,41). Thus, it seems that regions underlying binding capabilities are mainly present in the FAT domain, thus supporting the idea that events linked to FAK tyrosine phosphorylation activity determine kinetics of FAK and FRNK residency at FAs. Both FAT and Tyr 397 are required for complete tyrosine phosphorylation of downstream FAK targets like paxillin (20,21,41). Formation of Src-kinase/FAK complexes allows subsequent trans/cis-phosphorylation of other FAK tyrosine residues, inducing maximal FAK tyrosine kinase activ-ity (42) and direct or indirect recruitment of FA components (16). Like others (43), we found that transfection of FRNK leads to reduced phosphorylation of endogenous FAK at FAs, as well as reduced global FA tyrosine phosphorylation, probably because of decreased tyrosine phosphorylation of paxillin, p130cas, and tensin (20,21,44). Therefore, the faster recovery half-time of FRNK-Ycam at FAs may result from either a reduction in direct docking of FRNK at FA-localized proteins interacting with Tyr 397 or a decreased amount of FA-phosphotyrosines and, hence, a reduced number of potential interaction partners (43,45).
In support of intrinsic differences in FA docking, a C-terminal domain mutation which disrupts paxillin binding to FAK and FRNK affects only FA localization of FRNK but not FAK targeting (22,30,37). On the other hand, even if the N-terminal domain is unnecessary for complete integrin-dependent kinase activity of FAK (46) or FAK-dependent tyrosine phosphorylation of paxillin (41), its interaction with integrin (39) may act to stabilize and extend FAK association at FAs. Indeed, the N-terminal domain of FAK has been demonstrated to be involved in the control of FAK activity (15,47,48).
Several reports demonstrate that migration speed is increased by Ca 2ϩ signaling, for example in astrocytoma (28), smooth muscle cells (49), neutrophils (10), and neurons (50). Here, we have directly linked local Ca 2ϩ variations to the migration process by showing that Ca 2ϩ -dependent FA disassembly is associated with temporally and spatially restricted Ca 2ϩ increases around individual disassembling FAs. The migration process is complex, involving the regulation of cell protrusion and retraction both in space and time (51), characteristics shared by the complex patterns of Ca 2ϩ signaling reported here (Fig. 1). In cerebellar granule cells, Ca 2ϩ -dependent migration is correlated with the amplitude and frequency of Ca 2ϩ spikes (50), which may regulate different steps during migration. Migrating fibroblasts in the later stages of wound healing exhibit an increasing gradient of free Ca 2ϩ from the front to the rear (2, 3), supporting a role for spatially restricted Ca 2ϩ variations in the disassembly of FAs. Indeed, this might represent a mechanism by which the cell controls its polarity during migration. Oscillatory Ca 2ϩ signaling may serve to prevent inappropriate FA disassembly, which may occur during sustained Ca 2ϩ elevation, given the massive FA disruption seen for ionomycin-challenged cells.
It is widely held that the frequency and amplitude of Ca 2ϩ signals allow activation of specific Ca 2ϩ targets (54 -57). High Ca 2ϩ triggers diffusion of integrins out of FAs in adherent fibroblasts (52). Consistently, Ca 2ϩ increases lead to calpain activation and the liberation of a fraction of ␤ 2 integrins, which are tethered to cytoskeletal components (53). The Ca 2ϩ targets involved in FA disassembly might be more sensitive to high levels of Ca 2ϩ , because only strong local Ca 2ϩ elevations appear effective in triggering FA disassembly (Fig. 2). However, further studies are needed to explain why all FAs do not disassemble in response to massive ionomycin-induced Ca 2ϩ increases, and to explain why FAs respond to a given Ca 2ϩ spike and not the previous ones. One possibility is that repetitive increases in Ca 2ϩ around individual FAs are necessary to raise permissive signals beyond a putative threshold, thereby triggering disassembly, as has been proposed previously (58) for repetitive targeting of microtubules to individual FAs (which also promotes disassembly). In any case, the data presented here do not permit us to identify precisely the molecular processes underlying the generation of Ca 2ϩ oscillations nor to conclude that FAs are the exclusive sites of the active Ca 2ϩ oscillators which initiate Ca 2ϩ wave propagation. However, in cultured astrocytes, endoplasmic reticulum-associated proteins  (calreticulin and type 2 IP 3 receptors) and mitochondria have been described as Ca 2ϩ wave-amplification sites (69).
Currently, the precise Ca 2ϩ targets underlying FA disassembly are unknown, but they potentially include calcineurin (5), myosin light chain kinase (8), calpain (6,33), calreticulin (7, 58 -60), and the FAK-related proline-rich tyrosine kinase 2 (Pyk2) (61)(62)(63). Ca 2ϩ -dependent changes in FAK tyrosine phosphorylation activity have been described (23)(24)(25), and we found that Ca 2ϩ elevation rapidly (Ͻ1 min) increases FAK Tyr 397 phosphorylation (12), whereas Ca 2ϩ buffering with BAPTA reduces FAK Tyr 397 phosphorylation at FAs. Because FAK localization to FAs and autocatalytic activation (15,40) are involved in the subsequent Src activation and recruitment that enhances FAK activity (42), and both FAK (18,19) and Src (64,65) activity are linked to FA turnover, FAK may well be one of the Ca 2ϩ targets involved in Ca 2ϩ -dependent FA disassembly. In support of this idea, our FRAP experiments show that when [Ca 2ϩ ] is high, the FA-associated state of FAK-Ycam is prolonged, and the immobile FAK-Ycam fraction at FAs increases. In contrast, neither the recovery half-time nor the immobile fraction of FRNK-Ycam at FAs were Ca 2ϩ -sensitive, which is consistent with tyrosine phosphorylation events being necessary for the effect of Ca 2ϩ on FAK molecular dynamics. The Ca 2ϩ -insensitive dynamics of FRNK-Ycam strongly support the idea that the Ycam tag does not confer Ca 2ϩ -sensitivity to FAK-Ycam dynamics; rather, they support the idea that Ca 2ϩ affects FAK dynamics in a Ycam-independent manner. In further support of this, note that both endogenous FAK and FAK-Ycam in our cells undergo apparently normal Tyr 397 autophosphorylation, which is increased by [Ca 2ϩ ] elevation, and that native and FAK-Ycam-transfected U87 cells have similar rates of migration (12). We directly verified that Ca 2ϩ -induced conformational changes of Ycam do not underlie the slower dynamics of FAK-Ycam at FAs after ionomycin treatment (Fig.  5) by showing that FAK-GFP recovery half-times were also significantly increased when intracellular Ca 2ϩ levels were elevated (Fig. 7).
We propose that extended Ca 2ϩ -sensitive docking of FAK at FAs favors the turnover of FAs by increasing the probability of affecting FAK targets involved in FA disassembly, for example, RhoA (19) and Src (65). Consistently, extended interaction of ␣ 4 integrin subunits with paxillin is associated with faster FAK activation and increases in FA turnover and migration speed (66). Also, the inability of Pyk2 to maintain migration in FAKdeficient fibroblasts (67) does not involve differences in signaling pathways (16). Rather, the perinuclear distribution of Pyk2 (38), which, unlike FAK, is preferentially localized at FAs, suggests a time-limited co-localization of Pyk2 with its FA-substrates. Indeed, in FAK-deficient fibroblasts, a fusion protein of the Nterminal domain of Pyk2 and the C-terminal domain of FAK promotes both FA localization and migration (68). These data support our hypothesis that an enhanced time residency of FAK at FAs is correlated with increased FA turnover. Given the increases in time residency of FAK-Ycam at FAs and in FAK autophosphorylation when Ca 2ϩ is elevated, we suggest that the prolonged association of FAK with FAs is related to its level of autophosphorylation. We do not know whether Ca 2ϩ modifies other FA components at FAs, leading to a longer residency of FAK at FAs and increased transphosphorylation (15). The inverse possibility is that Ca 2ϩ signaling enhances FAK autophosphorylation at FAs, Tyr 397 -phosphorylated FAK being able to associate longer with FAs. This may arise because of increased amounts of FA-phosphotyrosines leading to a greater number of potential interaction partners (43,45) or because of conformational changes induced by phosphorylation.
Our data show that FAK molecular dynamics are fast and tunable by Ca 2ϩ . This may help to explain how the cell is capable of rapidly regulating FA stability during cell spreading and migration, in response to both local (FA) and remote (cytosolic) signals. Indeed, the fast flux of FAK between FA and cytosolic compartments raises the possibility that FA-associated FAK may be replaced by FAK having modified activity coming from the cytosol. Finally, our results indicate that regulation of FAK signaling might be achieved by changes in its association time with FA partner components.