Paxillin-dependent stimulation of microtubule catastrophes at focal adhesion sites

The dynamics of microtubules in higher eukaryotic cells are complex, being tightly regulated by a variety of microtubule-associated proteins. However, the precise mechanisms of this regulation are only partially understood. It has been well accepted that microtubules demonstrate random transitions from periods of growth to shortening and back at their plus-ends, termed `dynamic instability' (Mitchison and Kirschner, 1984), with most minus-ends being blocked owing to their association with microtubule organizing centers (MTOCs). Plus-end dynamics can be described by five parameters: the rates of growth and shortening, catastrophe frequency (transitions from growth to shortening), rescue frequency (transitions from shortening to growth) and time spent in pauses.

The overall pattern of microtubule dynamics differs between different cell types (Shelden and Wadsworth, 1993) and depends on the availability of serum factors (Danowski, 1998), as well as the phase in the cell cycle (Salmon et al., 1984). Most important, regional microtubule dynamics can differ within a single polarized cell. Thus, de-tyrosinated stable microtubules are found specifically in advancing lamellae of fibroblasts in wound-healing assays (Nagasaki et al., 1992). Dynamic microtubules in the rear of motile fibroblasts undergo more catastrophes than those in the front, whereas certain microtubules in lamellae, so-called `pioneer' microtubules, are characterized by almost no catastrophes (Wadsworth, 1999; Waterman-Storer and Salmon, 1997).

Microtubule dynamics in vivo can be specifically regulated by diverse protein factors (Amos and Schlieper, 2005; Howard and Hyman, 2007), certain posttranslational tubulin modifications (Tran et al., 2007) or tensile forces (Kaverina et al., 2002). In the past few years, more and more data indicate that interactions with the actin cytoskeleton play a crucial role in the local regulation of microtubule dynamics (Etienne-Manneville, 2004; Rodriguez et al., 2003). Indeed, it is the actin system and its regulators that possess region-specific differences in motile cells. Noteworthy in this regard, actin anchorage sites – focal adhesions – are able to precisely influence the dynamics of individual microtubules in the course of microtubule targeting of focal adhesions (Kaverina et al., 1998). The mechanism of targeting includes directional growth of the plus-end to the adhesion site and a short association of the microtubule tip with the adhesion plaque. Three subsequent scenarios are possible: first, the microtubule continues to grow, second it pauses in the adhesion and, third, it undergoes catastrophe and shrinks (Small and Kaverina, 2003). In this study, we show that catastrophe events are specifically enriched at focal adhesion sites. Such precise catastrophe activity has not been described before at any specific locus of interphase cells.

The nature of microtubule catastrophes can be diverse. They can be a part of dynamic instability and depend on the availability of free tubulin. Additionally, in vitro catastrophe frequency can be enhanced when microtubules polymerize against a stiff obstacle (Janson et al., 2003). Thus, a possibility exists that the catastrophes at adhesion sites are enhanced passively owing to the mechanical stiffness of adhesions or owing to the spatially restricted availability of free tubulin at dense adhesion plaques. However, it is known that catastrophes in cells are often induced by specific molecular factors, including stathmin or proteins of the kinesin-13 family (Cassimeris, 2002; Wordeman, 2005). Our present data suggest that a biochemical trigger locally activated at the focal adhesion plaque is crucial for catastrophe stimulation, whereas the physical influence of a dense adhesion structure makes a minor contribution to this process. In order to clarify what makes the adhesion plaque a preferred site for induction of catastrophe, we investigated focal adhesion molecules potentially involved in interacting with microtubules. In particular, we studied the focal adhesion scaffolding protein paxillin, which previously was found to bind to tubulin (Herreros et al., 2000).

Paxillin is a 68-kDa scaffold protein that contains many protein-binding modules and interacts with a variety of structural and signalling molecules (Brown and Turner, 2004). Paxillin contains two principle structural domains: five LD domains at the amino-terminus and four LIM domains at the carboxyl terminus. A number of kinases and phosphatases crucial for adhesion signalling bind to either LD or LIM domains, making paxillin a central player in several regulatory pathways (Brown and Turner, 2004). The paxillin LIM-domain region has been shown to serve as an anchor to a plasma membrane component and thus assure localization to focal adhesion sites (Brown et al., 1996), although the protein that mediates this association has not yet been identified. Of note, paxillin LIM2-LIM3 domains are also capable of binding to tubulin (Brown and Turner, 2002; Herreros et al., 2000).

In the present study, we show that the same two LIM domains of paxillin significantly change microtubule dynamics at adhesion sites when microinjected into fibroblasts. Substituting full-length paxillin at focal adhesions, LIM2-LIM3 protein reduces the catastrophe frequency exclusively at these locations, thus implicating paxillin in the local regulation of microtubule catastrophes.

In order to visualize microtubule dynamics clearly in relation to focal adhesions, we applied total internal reflectance fluorescence (TIRF) microscopy to illuminate only the ventral layer of cells attached to a glass substrate (Axelrod, 1989). Fish fibroblasts were transfected with mCherry-tubulin or 3xGFP-EMTB to visualize microtubules and GFP-paxillin or mCherry-paxillin to mark adhesion sites (Fig. 1A-C). Under TIRF conditions, we assumed that colocalization of a microtubule with an adhesion site indicates their close physical interaction. In live-cell time-lapse sequences, we observed that, when growing microtubules reached an adhesion site, they frequently underwent catastrophe and shrank. Individual microtubules often underwent multiple catastrophe events at a focal adhesion followed by subsequent rescues (Fig. 1A), whereas only few catastrophes occurred in adhesion-free areas (Fig. 1B). Analysis of overall adhesion numbers showed that approximately 40% of catastrophes occurred at adhesion sites, 12% at the cell edge and 48% elsewhere (Fig. 1D). As the overall adhesion area is considerably smaller than the adhesion-free ventral cell surface, these numbers indicate that adhesions serve as preferential sites for catastrophes.

A simple quantification of catastrophe events within a large area is insufficient for the analysis of the induction of microtubule catastrophes at local sites. The reason is that, because catastrophe is an event in which a microtubule switches from growth to shrinkage, it can occur only to a growing but not to a shortening plus-end. Thus, only those sites where growing microtubule ends were present were taken into consideration. Accordingly, we measured the length of elongation for microtubules while growing through adhesions or through an adhesion-free area. To obtain average persistent microtubule elongation, we normalized the average microtubule elongation by the average number of catastrophe events for each group of data. Strikingly, persistent microtubule elongation without catastrophes was seven times lower at adhesion sites (Fig. 1E) than elsewhere. These data suggest a specific mechanism that triggers microtubule catastrophes at adhesion sites.

A crucial question is whether catastrophes at adhesions are triggered by a specific biochemical mechanism or as a consequence of the increased density of actin filaments at the adhesion plaque. Such nonspecific mechanisms could include mechanical restraint or spatial restriction for free tubulin penetration. In order to answer this question, we analyzed the dynamics of mCherry-EB3-marked plus-tips of microtubules in the vicinity of GFP-paxillin-containing focal adhesion sites (Fig. 2).

Kymographic analysis of the dynamics of microtubule tips revealed two scenarios: either EB3 was released from a microtubule at an adhesion site, indicative of a catastrophe (Fig. 2A), or a growing tip proceeded through an adhesion (Fig. 2B). In the second case, the velocity of microtubule tips was notably reduced while moving through the adhesion site (Fig. 2B). In adhesion-free cytoplasm, ∼70% of microtubules grew at a rate of 0.1 μm/second or more (`fast microtubules') and ∼30% of microtubules grew more slowly (`slow microtubules'). At focal adhesions, the proportion of slow microtubules increased by up to 50% within the same microtubule population (Fig. 2D). At the same time, the catastrophe ratio dramatically increased at adhesion sites. For microtubules approaching adhesions, the probability of catastrophe increased from less than 5% on the approach route to ∼25% at the adhesion (Fig. 2F).

Combined, these data reveal a correlation between microtubule tip speed and catastrophe ratio. The percentage of `slow MTs' increases (Fig. 2D, red line) at the same location at inner adhesions (zones 0-1) where catastrophes are induced (Fig. 2F, blue line). These data suggest that the dense focal adhesion structure suppresses microtubule growth and thus increases the probability of catastrophe. This mechanism, however, is not specific for adhesions. We found that the catastrophe ratio of individual microtubules correlated inversely with their plus-tip velocity, both for microtubules growing through adhesions and through adhesion-free areas (Fig. 2E), suggesting that the microtubule catastrophe rate can be increased by cytoplasmic rigidity or spatially restricted concentration of tubulin dimers equally at adhesion plaques and at other locations. Strikingly, for microtubules with the same growth rate, the catastrophe ratio was higher at adhesion sites than elsewhere (Fig. 2E).

We have further tested whether softening of the adhesion structure decreases the ability of adhesions to induce catastrophes. For this purpose, we plated cells on a soft gelatin gel cushion (Seals et al., 2005), where strong tensile forces cannot be developed (Discher et al., 2005). Adhesions formed under these conditions were smaller than those on glass substrates (<2 μm versus <4 μm in length), but the catastrophe ratio at such adhesions was similar (23% on glass and 32% on gelatin gel). Together, these data indicate that a specific trigger of biochemical catastrophe acts at adhesion sites.

In line with this finding, some microtubules continued to polymerize when their plus-ends were apparently prevented from advancing by an obstacle. The polymerizing plus-ends of such microtubules remained immobile while the length of the microtubule increased, resulting in loop formation (supplementary material Fig. S1). Thus, mechanical obstacles do not necessarily cause microtubule catastrophe. Combined, our results strongly suggest that a biochemical mechanism, rather than a mechanical mechanism or spatially restricted concentration of tubulin dimers, drives induction of microtubule catastrophe at focal adhesion sites.

Having determined that a specific biochemical mechanism at the adhesion sites enhances microtubule catastrophe activity, we aimed to identify focal adhesion proteins potentially involved in the process. As focal adhesion composition and signalling changes during their life course, we first investigated whether induction of catastrophe depends of the adhesion maturation stage. To distinguish initial (early) focal adhesions from mature (late) adhesion sites, we used two adhesion markers: paxillin, which is present in the majority of adhesion sites, and zyxin, which has been shown to incorporate into adhesions at a later maturation stage (Zaidel-Bar et al., 2003). Simultaneous transfection of the microtubule marker GFP-EMTB, mCherry-paxillin and Cerulean-zyxin allowed us to compare microtubule dynamics at early (containing only paxillin) versus late (containing paxillin and zyxin) adhesions (Fig. 3A,B).

Overall quantification based on time-lapse live-image sequences revealed that ∼90% of adhesion-associated catastrophes occur at late adhesions, and only ∼10% at early adhesions (Fig. 3C). Nevertheless, taking into account that late adhesions are approached by a significantly higher number of microtubules, the final analysis showed that both types of adhesions were equally effective in catastrophe induction: 90% of approaching microtubules underwent catastrophe at adhesions (Fig. 3D) regardless of their maturation stage.

Although both zyxin and paxillin are present in late adhesions, we found that these adhesion proteins are distributed unevenly within single elongated mature adhesion sites: paxillin extends more distantly towards the cell edge than zyxin (Fig. 4). Such uneven distribution was found both at adhesions behind the leading edge (Fig. 4B) and sliding adhesions at the trailing edge (Fig. 4C) of the cell. We observed that a considerable number of microtubules underwent catastrophe at the distal ends of late adhesions (Fig. 1B). Detailed analysis of these catastrophes showed that they occurred at paxillin-rich adhesion domains devoid of zyxin (Fig. 4D-G). These data suggest that paxillin-associated adhesion structures are sufficient for induction of microtubule catastrophe.

We further examined the involvement of paxillin in the regulation of catastrophe by displacement of paxillin from adhesion sites. For this purpose, individual LIM domains of paxillin, including LIM2, LIM3 and tandem LIM2-LIM3 domains were produced as GST-fusion proteins (see Materials and Methods) and microinjected into cells. One hour after injection, cells were fixed and processed for immunostaining. Single GST-LIM2 (Fig. 5A,B) and GST-LIM3 (data not shown) showed mostly diffuse staining in the cytoplasm, with only occasional weak focal adhesion localization, whereas GST-LIM2-LIM3 was localized robustly to focal adhesions (Fig. 5C,D). Moreover, live-cell imaging of GFP-paxillin-expressing cells showed that injected LIM2-LIM3 domains caused displacement of GFP-paxillin from adhesion sites within 45 minutes (Fig. 5E,F). Immunostaining of injected cells confirmed paxillin displacement (data not shown) while another core adhesion component, vinculin, remained in the adhesions (Fig. 5D).

Notably, although LIM2-LIM3 domains have been shown to bind to tubulin (Brown and Turner, 2002; Herreros et al., 2000), no detectable localization to microtubules has been observed for the injected protein (Fig. 5C). Thus, if the LIM2-LIM3 domains of paxillin interact with microtubules in vivo, it can only occur in the vicinity of focal adhesions.

In order to determine whether substitution of paxillin by the LIM2-LIM3 domain at adhesion sites modulates microtubules at this location, we microinjected LIM2-LIM3 into fish fibroblasts co-transfected with GFP–β-tubulin and dsRed-zyxin. Time-lapse live-image sequences were recorded before the injection as well as 5 minutes after the injection. We found first that microtubules more often exhibited persistent growth resulting in an increase of microtubule length and density at the cell periphery (Fig. 6A,B). Analysis of microtubule dynamics revealed that, for each cell, the percentage of catastrophe events in the adhesion sites was 40% lower than in the same cell before injection (Fig. 6G,H). Typically, microtubules grew processively through three to four focal adhesions during a single growth phase in comparison with one to two adhesions in the same cell before injection (Fig. 6C-F). Along with a statistically significant decrease of catastrophe frequency at adhesions, no changes were observed in microtubule catastrophes in adhesion-free cytoplasm (Fig. 6G,H). This effect appeared to be specific for LIM2-LIM3 domains as microinjections of the paxillin LIM1 domain alone and either LIM1-LIM2 or LIM3-LIM4 domains had no influence on microtubule dynamics (data not shown).

Furthermore, we found that injection of LIM2-LIM3 paxillin domains in GFP-EB1-expressing cells resulted in an increased number of growing microtubules but had no effect on the velocity of microtubule growth. In particular, the number of EB1 comets indicating polymerizing microtubule plus-tips was increased by 30% 5 minutes after LIM2-LIM3 microinjection (supplementary material Fig. S2). However, the rate of microtubule growth did not change significantly (supplementary material Fig. S2), indicating that, other than decreasing the rate of catastrophe at the adhesion site, injection of LIM2-LIM3 protein had no detectable effect on microtubule polymerization.

Substrate adhesion turnover in fibroblasts and other cells is influenced by changes in microtubule dynamics (Small and Kaverina, 2003; Wittmann and Waterman-Storer, 2001). Our previous findings suggested that the origin of this interdependence is a direct cross-talk between microtubules and focal adhesions, involving their mutual interaction during dynamic targeting events (Kaverina et al., 2002; Kaverina et al., 1999). However, the molecular players involved in interactions between microtubules and focal adhesions remained unclear. Here, we have shown that microtubule catastrophes are specifically enriched at adhesion sites. In adhesion-free cytoplasm, a microtubule can grow for 4.9 μm on average without catastrophes, whereas, at adhesion sites, catastrophes can occur after only 0.7 μm of microtubule extension.

The accumulation of catastrophe events at the focal adhesion sites is notable for two reasons. First, focal adhesions serve as platforms where diverse signalling pathways are initiated, including those controlling cell motility and morphology, as well as proliferation and differentiation (DeMali et al., 2003; Martin, 2003). Second, microtubule catastrophe leads to a local release of a large group of microtubule-associated regulatory proteins. For example, microtubule catastrophe and disassembly results in the release of the microtubule-binding factor Rho GEF-H1 (Krendel et al., 2002) that mediates cross-talk between microtubules and the actin cytoskeleton through activation of Rho. Release of this factor from microtubules drives cell movements during the process of convergent extension of Xenopus laevis embryos (Kwan and Kirschner, 2005), making catastrophes crucial regulatory events in Xenopus development. Additionally, a set of plus-tip-binding proteins that are concentrated at the growing microtubule plus-ends is released at catastrophe sites. These proteins include adenomatous polyposis coli (APC) that is able to bind to members of the Rac- and Cdc42-regulatory pathways ASEF and IQGAP1 (Kawasaki et al., 2000; Watanabe et al., 2004). Thus, catastrophes might be important for modulation of Rac and Cdc42 signalling, controlling the actin cytoskeleton and cell polarity. Moreover, the release of APC can induce degradation of β-catenin and thus cause silencing of Wnt signalling (Polakis, 2007), suggesting that microtubule catastrophes might be involved in the regulation of cell proliferation and differentiation.

If signalling molecules are released from microtubules at adhesions, they could be brought into direct association with molecular factors concentrated at the adhesion plaques, including, among others, key players of the Src signalling pathway (Hanks et al., 2003; Parsons, 2003) and molecules controlling actin polymerization [PAK (Brown et al., 2002), Arp2/3 (Kaverina et al., 2003)]. The aim of this study was to understand the mechanisms whereby adhesions increase the probability of microtubule catastrophes. We addressed adhesion properties that might be responsible for enforcement of catastrophe.

Focal adhesions in cultured cells can be distinguished according to their molecular composition, dynamics and function (Zamir and Geiger, 2001). Early focal adhesions (also called focal complexes) are dot-like adhesions that assemble under the lamellipodium (Nobes and Hall, 1995; Rinnerthaler et al., 1988; Rottner et al., 1999). Within less than a minute of their formation, early adhesions either turn over or undergo a force-dependent transformation into late, or mature, focal adhesions (Zaidel-Bar et al., 2003). Late adhesions are considerably larger structures that are associated with actomyosin stress fibers and require contractility for their maintenance (Bershadsky et al., 2003; Galbraith et al., 2002). The protein composition of early and late adhesions is similar, although not identical. Paxillin, the major scaffolding adhesion factor, is one of the first proteins incorporated into early adhesions and remains associated during their transformation into mature late adhesions (Zaidel-Bar et al., 2003), and the majority of adhesion proteins colocalize with paxillin throughout the cell. By contrast, zyxin has been shown to be absent consistently from early adhesions and to be associated only with late adhesions (Rottner et al., 2001; Zaidel-Bar et al., 2003). Zyxin is also one of the first proteins dissociated from disassembling adhesions (Rottner et al., 2001). Based on these facts, we assumed that a paxillin-associated adhesion core might be considered distinct from zyxin-containing structures, and we used paxillin only and the zyxin-paxillin combination as markers for early and late adhesions, respectively.

Early adhesions are associated with poorly developed actin arrays and their rigidity is low compared with that of stress fiber-bound late adhesions (Bershadsky et al., 2003; Chen et al., 2004). However, microtubules undergo catastrophes both at early and late adhesions with the same efficiency (Fig. 3D). Decreasing the adhesion rigidity by growing cells on a gelatin cushion also did not decrease their ability to cause microtubule catastrophes. Additionally, a microtubule tip hitting an intracellular obstacle often does not result in a catastrophe. Combined, these observations suggest that mechanical factors probably play a minor role in promoting catastrophes. Another option for passive induction of catastrophes would be the spatially decreased availability of free tubulin at dense adhesion plaques. However, the density probably varies in the same way as does rigidity between early and late adhesions (i.e. early adhesions are less dense) as well as in adhesions under low tension on a gelatin gel substrate. Thus, our observations indicate that a mechanical factor does not contribute significantly to induction of catastrophes at adhesions. The same line of evidence suggests that spatial restriction of the tubulin concentration at dense adhesions does not strongly influence induction of catastrophes. We conclude that the influence of adhesions on microtubule catastrophes is attributable to a specific biochemical signal.

Catastrophes frequently occur at zyxin-free early adhesions or at the distal end of late adhesions rich in paxillin but devoid of zyxin. This finding suggests that the catastrophe factor is coupled with paxillin-associated, rather than with zyxin-associated, structures. The association of zyxin with adhesions is promoted by actomyosin contractility. Zyxin not only localizes to mature stress-fiber-connected adhesions but it can bind to stress fibers themselves under tension (Yoshigi et al., 2005). As catastrophes tend to occur in zyxin-free loci, the presence of a putative microtubule catastrophe factor at adhesions is probably neither actin associated nor tension dependent. Most likely, paxillin itself or paxillin-associated factors are involved in the induction of microtubule catastrophes at adhesion sites. Strikingly, a recent study showed that early adhesions and the distal part of late adhesions are enriched in phosphorylated paxillin (Zaidel-Bar et al., 2007). As we describe the same adhesion domains as areas of preferential induction of catastrophes, one can speculate that catastrophes could be associated with phosphorylated paxillin-based protein complexes.

It has been shown previously that paxillin binding to tubulin occurs through the LIM2 and LIM3 paxillin domains and requires intact LIM zinc-finger structures (Brown et al., 2002). We found that paxillin LIM2-LIM3 domains microinjected into fibroblasts localized specifically to adhesion sites, replacing full-length paxillin. This led to a decreased frequency of microtubule catastrophes at this location. Based on these data, we suggest a model that acknowledges the role of paxillin as a scaffolding protein at the adhesion sites (Fig. 7). According to this model, paxillin serves as a docking site for the catastrophe factor, possibly through phosphotyrosine-containing domains in the amino terminus (Fig. 7-1). When a microtubule approaches an adhesion site, a paxillin-associated factor can trigger it to undergo catastrophe and depolymerization directly (Fig. 7-2A) or through activation of a microtubule-associated catastrophe-inducing protein (Fig. 7-2B). Injected LIM2-LIM3 domains replace full-length paxillin at adhesion sites, thereby potentially displacing the putative catastrophe factor (Fig. 7-3). As a result, microtubules do not undergo catastrophe at adhesion sites any more (Fig. 7-4). If, during adhesion targeting, a microtubule directly binds to paxillin by means of the paxillin LIM2-LIM3 domain (Brown and Turner, 2002), this might additionally promote catastrophe by bringing the microtubule and catastrophe factor into close proximity. However, such an interaction might not be necessary for induction of catastrophe. Similarly, the putative catastrophe factor might not bind to paxillin directly but be part of a paxillin-organized adhesion core.

The nature of the catastrophe factor involved in microtubule depolymerization at the adhesion sites remains to be clarified. One of the well-known catastrophe-inducing molecules, such as stathmin (Cassimeris, 2002) or a member of the kinesin-13 family, for example MCAK (Wordeman, 2005), could be engaged in this process. However, none of these proteins has been found concentrated at focal adhesions. Thus, it is more likely that upstream regulators of these proteins, but not the proteins themselves, are specifically recruited to the adhesion plaques and play the role of `catastrophe factors' described in our model (Fig. 7-2B). As both stathmin (Larsson et al., 1997) and kinesin 13 (Andrews et al., 2004; Ohi et al., 2004) are active in the de-phosphorylated form, an adhesion-associated phosphatase could serve such a role. Alternatively, a putative catastrophe factor at adhesions could act through removal of certain stabilizing factors from the microtubule tip, thus allowing catastrophe-inducing molecules to complete their function. In this case, catastrophes might be induced by MCAK that is found at the tips of polymerizing microtubule in an already activated form (Moore et al., 2005) and can be delivered to adhesion sites during the course of microtubule targeting of focal adhesions.

In conclusion, our findings attribute to focal adhesions a decisive role in local regulation of microtubule catastrophes. Paxillin is probably a major upstream regulator of catastrophe induction at focal adhesion sites, although the actual catastrophe factor remains to be identified in future studies.

Goldfish fin fibroblasts (line CAR, no. CCL71; American Type Culture Collection) were maintained in DMEM with nonessential amino acids, 25 mM HEPES and with 10% FBS at 27°C. For experiments, cells were plated onto coverslips or glass-bottomed dishes (MatTek) and coated with human serum fibronectin (BD Biosciences) for at least 48 hours. Fibronectin was coated onto coverslips or glass-bottomed dishes (50 μg/ml in PBS for 30 minutes at room temperature). Gelatin gel cushion was prepared essentially as described previously (Seals et al., 2005). In brief, 2.5% gelatin in PBS containing 2.5% sucrose was allowed to polymerize for 30 minutes on glass-bottomed dishes (300 μl per dish). Next, the gels were crosslinked by 0.5% glutaraldehyde for 45 minutes. Free aldehyde groups were blocked by 1 mg/ml NaBH₄. Finally, the gels were coated with fibronectin, as described above.

mCherry-EB3 and Cerulean-EB3, made by replacing GFP in a pEGFP-EB3 construct (kind gift of A. Akhmanova, Rotterdam, The Netherlands) with Cerulean (kind gift of D. W. Piston, Nashville, TN) or mCherry (kind gift of R. Tsien, San Diego, CA), EGFP-EB1 (kind gift of A. Akhmanova, Rotterdam), pEGFP-tubulin (kind gift of J. Wehland, Germany), mCherry-tubulin (kind gift of R. Tsien) and 3xGFP-EMTB [microtubule-binding domain of E-MAP-115 (ensconsin) fused to three copies of EGFP (Bulinski et al., 1999), a kind gift of J. C. Bulinski, New York, NY] was used for microtubule plus tips and microtubule visualization. Neither tubulin fusion proteins (Rusan et al., 2001) nor EMTB (Faire et al., 1999) change microtubule dynamics and are routinely used for in vivo studies. GFP-paxillin (West et al., 2001), mCherry-paxillin (kind gift of S. Hanks, Nashville, TN), mCherry-zyxin, made by replacing GFP in a pEGFP-zyxin construct (kind gift of J. Wehland) with mCherry, and Cerulean-zyxin, constructed by cloning zyxin from mCherry-Zyxin into a mCerulean-C1 vector, were used for visualization of adhesion sites. Subconfluent monolayer cultures on 30 mm Petri dishes were used for transfection. For each dish, the transfection mixture was prepared as follows: 1.5-2 μg total DNA and 12 μl of Superfect lipofection agent (Qiagen) were mixed in 200 μl of serum-free medium. After 30 minutes of incubation at room temperature, a further 1.3 ml of medium containing 5% serum was added. Cells were incubated in this mixture for 4 hours at 27°C and the medium then replaced by normal medium containing 10% serum. After 24 hours, cells were replated at a desired dilution onto coverslips or glass-bottomed dishes for microscopy.

Injections were performed with sterile Femtotips (Eppendorf) held in a Leitz Micromanipulator with a pressure supply from an Eppendorf Microinjector 5242. Cells were injected with a continuous outflow mode from the needle under a constant pressure of between 20 and 40 hPa.

Cy3-tubulin was kindly provided by F. Severin (Max Planck Institute, Dresden, Germany). It was stored at a concentration of 20 mg/ml in aliquots at –70°C. For microinjections, Cy3-tubulin aliquots were diluted 1:3 with Tris-acetate injection buffer (2 mM Tris-acetate, pH 7.0, 50 mM KCl and 0.1 mM DTE) and used on the same day. GST only, GST-LIM1, GST-LIM2, GST-LIM3 and GST-LIM2-LIM3 fusion recombinant proteins were expressed in Escherichia coli and purified as described previously (Brown et al., 1998). GST, GST-LIM1, GST-LIM2, GST-LIM3 and GST-LIM2-LIM3 were stored at a concentration of 1.6 mg/ml at –70°C.

TIRF live-cell videos were acquired on a Nikon TE2000E microscope equipped with a Perfect Focus System and Nikon TIRF2 System for TE2000 using a TIRF 100× 1.49 NA oil-immersion lens and back-illuminated EM-CCD camera Cascade 512B (Photometrics) driven by IPLab software (Scanalytics). A 40 mW argon laser (Melles Griot) and 10 mW DPSS laser 85YCA010 (Melles Griot) were used for excitation. A custom-made double-dichroic TIRF mirror and emission filters (Chroma) in a filter wheel (Ludl) were used. Three-channel movies were made with the TIRF setting as described above for two channels and wide-field fluorescence using a BrightLine-CFP 2432A filter cube (Semrock) for the third channel.

Cells were injected and observed in an open chamber at room temperature on an inverted microscope (Axiovert 135TV; Zeiss) equipped for epifluorescence and phase-contrast microscopy. Injections were performed at an objective magnification of 40× (1.3 NA Plan Neofluar) and video microscopy with a 100× 1.4 NA Plan-Apochroma objective. Double fluorescence was achieved by a GFP-RFP Pinkel filter set (Chroma) in a custom-made filter wheel. Tungsten lamps (100 W) were used for both transmitted and epi-illumination. Data were acquired with a back-illuminated, cooled CCD camera from Princeton Research Instruments driven by IPLabs software. Paxillin displacement from focal adhesions was observed by live TIRF imaging of GFP-paxillin-expressing cells, as described above.

Cells coexpressing mCherry-alpha-tubulin and GFP-paxillin where photobleached for 10 seconds with a 10 mW DPSS laser 85YCA010 (Melles Griot) by focusing laser light in the focal plane with a custom-made lens (Nikon) placed in the position of the filter cube. Two-channel movies were recorded using TIRF microscopy, as described above, for 2.5 minutes after bleaching.

For immunostaining, cells were extracted for 1 minute in 0.25% Triton X-100 in cytoskeleton buffer (10 mM MES, 150 mM NaCL, 5 mM EGTA, 5 mM glucose and 5 mM MgCl₂, pH 6.1) and fixed for 20 minutes in 3% paraformaldehyde in cytoskeleton buffer. Immunostaining was performed using polyclonal goat antibodies against glutathione S-transferase (GST) and mouse antibodies against vinculin.

Microtubule catastrophe: time-lapse movies of cells expressing 3xGFP-EMTB to label microtubules and mCherry-paxillin to mark adhesion sites (5 seconds/frame) were processed by background subtraction and intensity adjustment using ImageJ software. The coordinates of each catastrophe event were recorded and then checked to determine whether there was an adhesion site with the same coordinates.

Average microtubule elongation: time-lapse movies of cells expressing mCherry-tubulin and GFP-zyxin (5 seconds/frame) were processed by background subtraction and intensity adjustment using ImageJ software. The trajectories of microtubule plus-end movement were followed using a plug-in of ImageJ (`manual tracking'). Microtubule elongation in the cytoplasm or at adhesion sites was normalized by the number of catastrophes in each location for each microtubule. Then the average elongation distance was calculated for all microtubules. Statistical analysis was performed in Microsoft Excel.

Tip dynamics at adhesion sites: time-lapse movies of cells expressing mCherry-EB3 and GFP-paxillin (5 seconds/frame) were processed for intensity adjustment using ImageJ. For uniform analysis, only the most common cases when microtubule tips approached adhesions from their proximal ends were considered. Narrow areas around adhesion sites were delineated and combined into kymographs. The area was divided into 1-μm-long zones (Fig. 2C). For each microtubule tip, the time it spends in each zone and place of catastrophe was calculated manually. Microtubules that pass through the zone in 1-2 frames (10 seconds or less) were considered as fast microtubules and in three frames or more (>15 seconds) as slow microtubules. For each zone, the number of approaching microtubule tips and disappearing tips was calculated. For zones –1, 0 and 1, the probability of a microtubule undergoing catastrophe versus the time the microtubule tip spent in the zone was calculated. Statistical analysis was performed in Microsoft Excel.

Catastrophes at early and late adhesions: time-lapse movies of cells expressing 3xGFP-EMTB, mCherry-paxillin and Cerulean-zyxin (5 seconds/frame) were used. Trajectories of microtubule plus-end movement were followed using a plug-in (`manual tracking') of ImageJ software. The coordinates of each catastrophe were recorded and then checked against the localization of paxillin and zyxin by using ImageJ software. Statistical analysis was performed in Microsoft Excel.

Analysis of microtubule dynamics after microinjection: calculations of microtubule growth velocity and catastrophe events were made using MetaMorph software, and statistical analysis was performed in Microsoft Excel.

We thank Steve Hanks and Ethan Lee for helpful discussion and advice. This study was funded by Development funds of the Department of Cell and Developmental Biology, Vanderbilt University Medical Center and by a pilot project within grant ACS IRG-58-009-49 to I.K., by Austrian Science Fund grant #P16743-B09P to J.V.S. and NIH grant RO1 GM47607 to C.E.T.