Microtubule-targeting-dependent reorganization of filopodia

Cross-communication between the microtubule system and actin cytoskeleton is a basic phenomenon that underlies fundamental cell processes including migration, polarity and division (Goode et al., 2000; Kodama et al., 2004; Rodriguez et al., 2003). Cell migration is typified by lamellipodium protrusion, often consisting of a fine dendritic actin network between large actin filament bundles (Small et al., 2002), and requires coordination with adhesion site disassembly and cell body retraction (Ridley et al., 2003). When a migrating cell responds to an extracellular cue and turns, the actin cytoskeleton must concurrently reorganize. The microtubule system radiates out from the cell body and enters the lamellipodium where microtubule plus-ends may interact with actin target sites. An important feature of cross-talk between these two branches of the cytoskeleton is microtubule-directed reorganization of actin structure during cell guidance (Rodriguez et al., 2003).

In most cells, polymerization of microtubules is nucleated at the centrosome and proceeds toward the cell periphery. Microtubule orientation is polarized, the minus-end is pointed toward the centrosome and the plus-end is pointed toward the cell periphery (Dammermann et al., 2003). Disposition of microtubules in the lamellipodium of a migrating cell is dependent on multiple processes. Microtubules enter the lamellipodium by polymerization at the plus-end (Kabir et al., 2001); however, at the same time microtubules move rearward closely correlated with actin retrograde movement in the lamellum (Salmon et al., 2002). Global disruption of actin filaments enhances microtubule entry into the peripheral domain of growth cones while inhibition of myosin II interferes with filopodia formation and decreases microtubule entry (Zhou et al., 2002), indicating that actin architecture is a key factor. In addition, mechanical stress plays a role in adhesion site formation and microtubule entry into areas of actin protrusion. Local tensile stress applied to the lamellipodium of B16F1 cells induces formation of adhesion sites and enhances polymerization of microtubules towards the leading edge (Kaverina et al., 2002a).

After reaching the cell periphery, microtubule ends may interact with several target sites involved in coordination of cell movement. Microtubule targeting events in the periphery of fibroblasts precede adhesion site remodeling and dissociation (Kaverina et al., 1999), a process that is dependent on kinesin (Krylyshkina et al., 2002). More recently, findings have shown that focal adhesion disassembly after nocodazole washout is dependent on focal adhesion kinase and dynamin (Ezratty et al., 2005). Consistent with a role of microtubule targeting to focal adhesions in cell motility, nocodazole disrupts tail retraction and slows migration (Ballestrem et al., 2000). Microtubules align with actin arcs and filopodia in the peripheral domain of neuronal growth cones (Schaefer et al., 2002). Mediators bridging the interaction between microtubules and filopodia have not been conclusively identified and the consequence of this interaction on filopodia dynamics is not known. Microtubule plus-end tracking proteins are prime candidates for bridging this interaction because of their concentration at the tips of microtubules (Carvalho et al., 2003) and ability to associate with effectors of the actin cytoskeleton (Rodriguez et al., 2003). For example, CLIP170 binds to Cdc42 and Rac1 through IQGAP1 forming a tripartite complex. Disruption of this complex causes mislocalization of actin protrusion (Fukata et al., 2002). Another plus-end binding protein, EB1, binds to the formin mDia2 (Wen et al., 2004), an effector of filopodia formation (Schirenbeck et al., 2005).

In this study we developed an assay to quantify targeting of peripheral microtubules to filopodia, examined the effect of targeting on filopodia reorganization, and probed the molecular mechanism by which microtubules interact with filopodia. We identified microtubule targeting to filopodia in melanoma cells and fibroblasts. Targeting events were closely associated with filopodia movement and merging, and importantly, uncoupling targeting increased filopodia density in lamellipodia wings. Furthermore, we found that targeting events were independent of local adhesion sites and microtubule plus-end-binding proteins. We propose that microtubule-dependent regulation of filopodia merging is a mechanism for control of filopodia density and may reorganize actin architecture required for directed cell migration.

We developed an assay for quantifying microtubule targeting to filopodia based on pixel overlap of these cytoskeleton structures in digital immunofluorescence images. As shown in Fig. 1A, individual microtubules and filopodia are easily distinguished making the images amenable to quantitative analysis. In our analysis, microspikes (actin bundles that do not protrude beyond the cell margin) are called `filopodia'. In the lamellipodium shown in Fig. 1A, there are five `peripheral' microtubules defined as those contacting or intersecting a line (baseline) connecting filopodial bases. A positive targeting event a `hit', was defined as a microtubule intersection with the base of a filopodium. From the two enlarged regions (Fig. 1B,C), three microtubules were scored as hits (+) on filopodia and two microtubules were scored as misses (-). To determine if the interaction between microtubules and filopodia exhibited a spatial preference, lamellipodia were divided into five equal zones and the number of microtubules, filopodia and targeting events for each zone was quantified. A disproportionately large number of targeting events, 64%, occurred in the lamellipodia `wings' (zones 3), whereas only 23% of the targeting events occurred in zones 2 and 13% in zone 1 (Fig. 1D). The preference for targeting events in lamellipodia wings is not explained by a biased distribution of microtubules or filopodia since the number of these structures was approximately equal for each zone (Fig. 1D). Targeting events were observed quite frequently and involved a significant proportion of the filopodia and peripheral microtubule population. The mean number of filopodia per cell ± the standard error of the mean (± s.e.m.) was 15.0±0.8 and 18.3% were targeted by microtubules. The mean number of peripheral microtubules per cell was 7.0±0.6 and 45.6% targeted filopodia. Similar to observations in neuronal growth cones (Schaefer et al., 2002), some microtubule ends did extend beyond the filopodium base (approximately 1 μm) and aligned parallel to the filopodium shaft (see Table 1); however, microtubules ends were not observed to reach filopodia tips.

We evaluated the possibility that the apparent targeting could be explained by random overlap of microtubule ends with filopodia. Since chance targeting may be fitted to a binomial distribution, we determined the cumulative probability of observing ≥45.6% positive microtubules (see Materials and Methods). Assuming random targeting, the cumulative probability of observing ≥45.6% positive microtubules is exceedingly small (P<1×10^-6), thus ruling out the possibility that our observations are explained by chance and necessitating a mechanism for directing microtubules to filopodia. In addition to B16F1 cells, we observed microtubule targeting to filopodia in mouse dermal fibroblasts (see below) and NIH 3T3 fibroblasts (data not shown) indicating the generality of this phenomenon.

We extended our studies to investigate the possible consequences of targeting by performing experiments in live B16F1 melanoma cells co-transfected with YFP-β-tubulin to visualize microtubules, and YFP-fascin to visualize filopodia. The examination focused on lamellipodia wings since we observed the highest frequency of targeting events in these regions. The video sequence Movie 1 (see supplementary material) shows turning of a targeted filopodium while the nontargeted filopodium remains about its original position. A filopodium (FL1) is targeted by a microtubule (MT1) beginning at 25 seconds and then targeted again by another microtubule (MT2) beginning at 60 seconds (Fig. 2A; see Movie 1 for complete sequence). Filopodium FL1 turns through 60° after targeting by microtubules, while the reference filopodium (FL2) remains about its original angle (Fig. 2C). The video sequence Movie 2 in supplementary material shows another event where the targeted filopodium turns and then merges with an adjacent filopodium. The targeted filopodium (FL1) turns, merges with a nearby filopodium (FL2), and the resultant filopodium (FL1 + FL2) continues to turn (Fig. 2B; see Movie 2 for complete sequence). FL1 rotates through 50° towards the lamellipodium midline; by contrast, the adjacent reference filopodium (FL3) remains in its initial orientation (Fig. 2D). We examined 15 targeting events in lamellipodia wings of 12 different cells and 13 of the events resulted in filopodia turning. One of the two exceptions, in which no filopodia turning was observed after targeting, was the filopodium between FL2 and FL3 (Fig. 2B). The average duration of a targeting event (±s.e.m.) was 52±9.5 seconds and the difference in angle between targeted and reference filopodia pairs increased progressively during targeting (Fig. 2E). 80% of all targeting events resulted in merging with a nearby filopodium after 60 seconds (Fig. 2F). The targeting events that did not result in merging were probably due to lower filopodia density in some cells. Furthermore, we observed movement of filopodia out of the focal plane during targeting events (Fig. 2B) suggesting that targeted filopodia are not restricted to lateral movement, but may also lift away from the substrate as a consequence of microtubule targeting. This idea was explicitly examined in total internal fluorescence microscopy studies (see below).

We examined the effect of targeting on filopodia merging and filopodia density by uncoupling the microtubule-filopodia interaction with nocodazole. Nocodazole treatment rapidly decreased the number of peripheral microtubules thus precluding any targeting events (Fig. 3A). At 1.0 minute, a time at which microtubule-filopodia interaction was completely uncoupled, the number of filopodia merging events in zones 3 was decreased (Fig. 3B). By contrast, uncoupling targeting increased filopodia density in zones 3 with a time course that closely paralleled the nocodazole effect on filopodia merging (Fig. 3C). These results show that microtubule targeting in lamellipodia wings correlates with filopodia turning and merging, and moreover, suggest a unique mechanism for control of filopodia number.

In total internal reflection fluorescence (TIRF) microscopy the evanescent wave decays exponentially, penetrating less than 150 nm beyond the reflecting interface, making this technique ideal for examining processes that occur near the ventral cell surface (Krylyshkina et al., 2003; Steyer and Almers, 2001; Truskey et al., 1992). We used this technique to determine the spatial proximity between microtubules or filopodia and the ventral cell surface during targeting events. Fig. 4A compares wide-field and TIRF images of one nontargeted filopodium (asterisk) and three targeted filopodia (arrowheads). Most microtubule ends detected in wide field were also clearly visible by TIRF and therefore near the ventral surface of the cell (Fig. 4A). Overall, 81% of targeting microtubules and 77% of nontargeting microtubules were visible by TIRF microscopy (Fig. 4B). This indicates that microtubules tend to be proximal to the ventral surface independent of targeting to filopodia. Of the four filopodia visible in wide field, two were visible in TIRF, one of which was targeted and one of which was not. The two other targeted filopodia were not visible in TIRF (Fig. 4A). Overall, 40% of targeted filopodia, in contrast to 65% of nontargeted filopodia, were visible by TIRF microscopy (Fig. 4B). The lower proportion of targeted filopodia visible in TIRF suggests that movement of filopodia away from the ventral surface is the result of microtubule targeting and after targeting; microtubules remain near the ventral surface. These results are supported by TIRF microscopy of live cells co-transfected with YFP-tubulin and YFP-fascin. As shown in the wide-field sequence, a microtubule (MT1) targets a filopodium (FL3) and another microtubule (MT2) targets another filopodium (FL2), and as previously observed, the targeted filopodia turn toward the lamellipodium midline (Fig. 4C, wide field). By sequential imaging, the same two targeting events were also observed by TIRF microscopy. The two targeted filopodia, FL2 and FL3, disappear after microtubule targeting indicating upward movement out of the evanescent field and away from the ventral cell surface. By contrast, the nontargeted filopodium (FL1) remains visible through the TIRF sequence (Fig. 4C, tirf).

In the TIRF microscopy studies we observed microtubules below filopodia during targeting events. Thus, a possible explanation for targeting is focal adhesions, which if present at the bases of filopodia may be the true target site and mechanistically explain the interaction between microtubules and filopodia. We addressed this question by examining localization of adhesion markers in the lamellipodia of cells displaying microtubule targeting. We chose markers present throughout adhesion development from early sites (focal complexes) to more mature sites (focal adhesions) (Zaidel-Bar et al., 2003; Zaidel-Bar et al., 2004). Focal adhesion kinase (FAK) and paxillin were found to be present behind the lamellipodium margin at adhesion sites as expected (Fig. 5A,B). A third marker, vasodilator-stimulated phosphoprotein (VASP) localizes to filopodia tips and the lamellipodium leading edge in addition to adhesion sites (Fig. 5C). We quantified the percentage of targeted and nontargeted filopodia bases that were positive for these three adhesion site markers. Localization of FAK to filopodia bases was relatively rare, about 3%. We observed greater localization of paxillin and VASP, about 17% and 22%, respectively. Most importantly, the percentage of targeted filopodia bases positive for adhesion site markers equaled that of nontargeted filopodia bases (Fig. 5). From the lack of correlation between adhesion markers and targeting events, we conclude that microtubule targeting to filopodia is not dependent on the local presence of focal complexes or focal adhesions.

CLIP170 is a plus-end tracking protein that associates with polymerizing plus-ends of microtubules for a short time and then dissociates behind the region of most recent growth (Perez et al., 1999). This property of CLIP170 was used to determine the polymerization state of microtubules during targeting events. A time-lapse series of cells co-transfected with GFP-CLIP170 and mRFP-actin was acquired by dual-channel simultaneous fluorescence microscopy. Fig. 6 illustrates a polymerizing microtubule plus-end, highlighted by comet-shaped GFP-CLIP170 fluorescence approaching a filopodium base and then rapidly turning to align with the filopodium. This kinetic behavior indicates that the trajectory of growing microtubules is affected by proximity to the filopodium base and demonstrates a close spatial context where plus-end proteins could bridge interaction between microtubules and filopodia.

We tested the possibility that plus-end tracking proteins are required for the targeting interaction between microtubules and filopodia, by using a small hairpin RNA interference approach in B16F1 cells. The interference constructs encode a small hairpin structure and soluble GFP marker, thereby allowing easy identification of knockdown cells by GFP fluorescence (Kojima et al., 2004). In addition, we extended our studies to include dual knockout mouse dermal fibroblasts (MDF) deficient in both CLIP170 and CLIP115. We assessed protein levels in these cells by immunostaining with polyclonal antibodies reactive with the head domain of CLIP170 and CLIP115 (Hoogenraad et al., 2000) and antibodies specific for end-binding (EB) protein isoforms (Komarova et al., 2005). Through RNA interference we achieved 87% knock down of CLIP115 and CLIP170 proteins in B16F1 cells (CLIP KD in Fig. S1A,B in supplementary material). As expected, there were no detectable levels of CLIP protein in the dual knockout MDF cells (CLIP KO in Fig. S1A,B in supplementary material). In B16F1 cells, we obtained 92%, 81% and 76% knock down of individual EB1, EB2 and EB3, respectively. Because of functional redundancy (Komarova et al., 2005), we co-transfected B16F1 cells with the EB1 and EB3 knock down constructs to achieve simultaneous depletion. In dual EB1+EB3 knockdown cells, protein levels were reduced by 73% and 86%, respectively (Fig. S1C,D in supplementary material). We next analyzed targeting events in these cells depleted of microtubule plus-end-binding proteins. Fig. S2A in supplementary material shows combined images of microtubules and actin in B16F1 CLIP KD cells and MDF CLIP KO cells. The arrowheads indicate targeting of microtubules to filopodia. We observed normal filopodia targeting frequency in B16F1 CLIP KD cells compared to nontransfected control cells (Fig. S2C in supplementary material). In agreement with the experiments performed in B16F1 cells, we observed no effect of dual CLIP knock out on targeting frequency in MDF cells. Likewise, depletion of EB proteins did not affect filopodia targeting frequency (Fig. S2B,D in supplementary material). These results indicate that the plus-end-binding proteins are not required for microtubule targeting to filopodia.

Microtubules undergo periods of sustained growth and shortening punctuated by transitions from growth to shortening, called catastrophe, or shortening to growth, called rescue. This behavior of microtubules is referred to as dynamic instability (Mitchison and Kirschner, 1984) and is a supposed mechanism by which microtubule ends probe the cytoplasm for target sites. We examined targeting under conditions where microtubule dynamics are dramatically altered by the compound taxol. Treatment of cells with taxol for 30 seconds caused a burst of microtubule growth into the lamellipodium periphery after which time the number of peripheral microtubules rapidly declined as the cell leading edge continued to move forward (Fig. 7A). As expected, brief treatment with taxol increased the number of peripheral microtubules (Fig. 7B) and increased the proportion of filopodia targeted (Fig. 7C). Although the number of peripheral microtubules increased, the proportion of microtubules targeted to filopodia did not change (Fig. 7D). These findings indicate that forcing all microtubules into a single dynamic state, the growth phase, is sufficient to induce microtubule targeting to filopodia and furthermore, suggests a mechanism by which microtubules find their target at short ranges without requirement of dynamic instability.

Cross-communication between the microtubule system and actin cytoskeleton has emerged as a key interaction underlying fundamental cell processes ranging from motility and growth cone guidance to cell division (Rodriguez et al., 2003). One goal of our studies was to gain understanding of how the interaction between microtubules and a specific actin cytoskeleton structure, filopodia, plays a role in cell motility. We developed a quantitative microtubule targeting assay which allowed investigation into the mechanism and function of targeting. From studies using mouse melanoma cells and fibroblasts we gained insight into the mechanism of targeting and most importantly, found close temporal correlation between targeting events and filopodia movement, merging and control of filopodia density.

B16F1 melanoma cells plated on laminin substrate exhibit protruding lamellipodia and entry of microtubules into the protrusion that reach the bases of filopodia. Using a digital fluorescence image-based approach, we provided a detailed analysis of microtubule targeting to filopodia in these cells. Targeting events occur frequently, in fact, far beyond frequencies that are explained by chance, involving a significant proportion of both peripheral microtubules and filopodia, suggesting requirement of a molecular mechanism to mediate this interaction. Microtubules enter the peripheral domain of neuron growth cones and then may align parallel to filopodia (Schaefer et al., 2002). Similarly, we observed microtubule tips traveling beyond the base and aligning with the filopodium shaft; however, microtubules tips reaching the ends of filopodia were not observed. On average, peripheral microtubules and filopodia are evenly distributed throughout the width of lamellipodia, but more targeting events occur in lamellipodia wings. Thus, in an effort to investigate the function of this targeting interaction we focused our attention on filopodia reorganization in the wings of lamellipodia.

We have previously proposed a model that describes a mechanism for formation of filopodia, termed the convergent elongation model (Svitkina et al., 2003). Filopodia initiation begins by reorganization of the actin dendritic network and collision of polymerizing actin barbed ends. Long parallel actin bundles form through continued polymerization and stabilization by the actin crosslinker, fascin (Edwards and Bryan, 1995), forming nascent filopodia or microspikes (Svitkina et al., 2003). Although this mechanism accounts for the process of filopodia formation, we lack understanding of how mature filopodia distribution is regulated and how this may relate to cell movement. In our current studies, we show a strong correlation between microtubule targeting and later filopodia kinetic events, namely, filopodia turning and merging. Moreover, TIRF microscopy studies revealed that filopodia are not restricted to lateral movement, but may also move upward, away from the ventral cell surface after targeting. Importantly, we showed that uncoupling the microtubule-filopodia interaction by brief treatment with nocodazole decreases merging events and increases filopodia density. This provides a unique microtubule-dependent mechanism for control of filopodia reorganization in lamellipodia wings where the transition from actin network protrusion to retraction must occur. When a cell turns toward a stimulus it must retract on the opposite side and reorganize the actin cytoskeleton (Kaverina et al., 2002b), in particular, by downregulating filopodia activity. Microtubules may participate in this process by mediating filopodia movement and merging which, in turn, may control filopodia density in lamellipodia during directed cell migration. Our hypothesized role of microtubule targeting to filopodia in lamellipodia turning is consistent with studies in neuronal growth cones. Microtubules align with filopodia in growth cones, and turning of growth cones is induced by local application of compounds that interfere with microtubule polymerization. Nocodazole causes depolymerization, local growth cone collapse and turning away from the side of application (Buck and Zheng, 2002). Taxol causes entry of microtubules into the peripheral domain and turning toward the side of application (Buck and Zheng, 2002).

We observed microtubule interaction with filopodia for periods longer than 1 minute during some targeting events. Two possible candidates for bridging this interaction are microtubule plus-end-binding proteins and molecular components of cell adhesion sites. Findings from the laboratory of J. V. Small show targeting of microtubule tips to adhesions sites in the cell periphery of fibroblasts (Kaverina et al., 1998). Targeting events precede focal adhesion dissociation and retraction of the cell edge (Kaverina et al., 1999). Using TIRF microscopy, we observed microtubule tips at the ventral cell surface sometimes below the bases of filopodia. These findings raised the possibility that microtubules may target focal adhesion sites, if present, at the bases of filopodia. Focal adhesion kinase, paxillin and VASP, markers of focal complexes and focal adhesions, were not detected at filopodia bases. Microtubule tips targeted to adhesion sites are resistant to depolymerization by nocodazole (Kaverina et al., 1998). We found that brief nocodazole treatment ablated peripheral microtubules and destroyed all targeting events. We thus concluded that filopodia are bona fide targets for microtubules not dependent on local presence of focal complexes or focal adhesions. We explored the second major possibility that plus-end-binding proteins may bridge the microtubule-filopodia interaction. CLIP170 and CLIP115 are two plus-end-binding proteins that interact with CLIP-associated protein (CLASP) to mediate microtubule organization (Akhmanova et al., 2001) and with the IQGAP1-Cdc42-Rac1 molecular complex to mediate microtubule capture (Fukata et al., 2002). We explored the possibility that these proteins may bridge the microtubule-filopodia interaction in RNA interference studies. Simultaneous knock down of CLIP170 and the functionally related CLIP115 did not affect targeting in B16F1 cells. Consistent with these results, we observed normal microtubule targeting to filopodia in dual knockout MDF cells deficient in CLIP170 and CLIP115. Furthermore, depletion of EB protein family members did not alter targeting. We concluded that microtubule plus-end-binding proteins are not required for targeting of microtubules to filopodia.

What is the mechanism by which microtubules find their target site? Microtubules possess an intrinsic property known as dynamic instability, in which their plus-ends alternate through phases of polymerization and depolymerization, punctuated by catastrophe and rescue transitions (Mitchison and Kirschner, 1984). Furthermore, microtubules exhibit asymmetric behavior, persistently growing toward the plasma membrane and then displaying frequent transitions between growth and shortening near the cell periphery (Komarova et al., 2002). We tested the hypothesis that peripheral microtubules require these frequent transitions in order to successfully target filopodia. Taxol switches all microtubules to the growth phase by slowing tubulin dissociation (Wilson et al., 1985), essentially halting dynamic instability, and the microtubules continue to grow until most of the tubulin is consumed in the polymerized state. Following treatment with taxol, we observed a burst in targeting activity indicating that the growth phase is sufficient for microtubules to successfully reach filopodia. Fluorescent speckle microscopy studies demonstrate correlated movement of microtubules with actin in protruding epithelial cells (Salmon et al., 2002) suggesting a physical link between these two cytoskeleton networks. Findings in neuron growth cones indicate that actin bundles are required for entry of microtubules into the peripheral domain (Zhou et al., 2002). Our results of the taxol experiments favor a mechanism in which a population of peripheral microtubules is guided to their target site, likely by actin bundles, and does not require random microtubule search and capture through dynamic instability.

Our studies have revealed a unique mechanism by which microtubules may directly participate in directed cell motility through control of filopodia dynamics, namely, filopodia merging and control of filopodia density. Microtubule interaction with actin bundles has been firmly established in growth cones. Here we have shown targeting of microtubules in B16F1 mouse melanoma cells and fibroblasts suggesting that this interaction is of broad biological importance. Study of microtubule targeting to filopodia in new cell types paves the way for future studies in which molecular biology approaches are more easily implemented and will in turn facilitate identification of molecular components mediating this interaction.

B16F1 mouse melanoma cells (American Type Culture Collection, Rockville, MD) were grown in DMEM (Gibco Invitrogen, Carlsbad, CA) supplemented with 10% fetal bovine serum (Atlanta Biologicals, Lawrenceville, GA), L-glutamine, pyridoxine hydrochloride and antibiotics. The CLIP170, CLIP115 dual knock mouse dermal fibroblast (MDF) cell line was produced and generously provided by Niels Galjart from the Department of Cell Biology and Genetics, Erasmus Medical Center, Rotterdam, The Netherlands. The single CLIP170 and CLIP115 knockout alleles have been described previously (Akhmanova et al., 2005; Hoogenraad et al., 2002). In mice, the genes encoding CLIP170 [Clip1 (Rsn)] and CLIP115 [Clip2 (Cyln2)] are on chromosome 5 approximately 10 megabases apart. Thus, generation of the CLIP170, CLIP115 dual knock mice requires meiotic recombination between the two single mutant alleles. For the CLIP170 and CLIP115 knockout strains 13% recombination efficiency was obtained out of 200 mice analyzed, consistent with the estimated distance between the two genes. Heterozygous dual knockout mice were mated and MDF cells were isolated from wild-type and homozygous dual knockout adult offspring. MDF cells were grown in the same medium as above except supplemented with 20% fetal bovine serum. Cells were maintained in a humidified, 10% CO₂ incubator at 37°C.

The GFP-CLIP170 construct contains cDNA encoding full length rat CLIP170 in the pEGFP-C1 vector. The monomeric RFP-actin (mRFP-actin) construct contains cDNA encoding for the human β-actin subunit in the pmRFP1-C1 vector. The YFP-fascin and YFP-tubulin constructs contain cDNA encoding for human fascin-1 and mouse β5-tubulin subunit in the pEYFP-C1 and pEYFP-N1 vectors, respectively. The GFP-paxillin construct was generously provided by A. F. Horwitz, University of Virginia, Charlottesville, VA. Human vasodilator-activated phosphoprotein cDNA was kindly provided by Frank B. Gertler, Massachusetts Institute of Technology, Cambridge, MA. B16F1 cells were transfected using Fugene 6 reagent (Roche Diagnostics, Indianapolis, IN) according to the protocol provided by the manufacturer. Experiments were performed 24-48 hours after transfection. For live-cell imaging studies, the transfected B16F1 cells were suspended in Leibovitz's L-15 medium (Gibco Invitrogen) without Phenol Red, supplemented with 10% fetal bovine serum, L-glutamine and antibiotics, and then plated onto laminin-coated coverslips.

The target sequences used in this study are as follows: CLIP115, CTGGAAATCCAAGCTGGAC; CLIP170, GGAGAAGCAGCAGCACATT; EB1, GCCTGGACCAGCAGAGCAA; EB2, GATGAATGTTGATAAGGTA; EB3, ACTATGATGGAAAGGATTAC. The EB3 target sequence was used in previous studies (Komarova et al., 2005).

For simultaneous knock down of CLIP115 and CLIP170 the target sequences were inserted in tandem into the pEGFP-C1 vector. The EB1 target sequence was inserted into the pGShin vector (Kojima et al., 2004). The EB2 and EB3 sequences were inserted in to the pECFP-C1 vector. B16F1 cells were transfected with the RNA interference constructs using the Fugene 6 reagent. The microtubule targeting assay was performed 3-4 days after transfection of the knockdown constructs. Protein knock down was confirmed by immunostaining with rabbit polyclonal anti-CLIP head domain antibodies (Hoogenraad et al., 2000) and monoclonal antibodies specific for EB proteins (Komarova et al., 2005).

B16F1 cells were plated at low density onto glass coverslips previously coated with 30 μg/ml laminin (Invitrogen Corporation) for 24 hours at 4°C and blocked with 2% bovine serum albumin. For total internal reflection fluorescence studies, high refractive index coverslips were used (Optical Analysis Corporation, Nashua, NH). Cells were incubated on the coverslips for 30 minutes at 37°C in 10% CO₂ to allow spreading and formation of lamellipodia and then fixed in phosphate-buffered saline (PBS) containing 1% glutaraldehyde and 1% Triton X-100 for 30 minutes at 22°C. Coverslips were washed in PBS, neutralized with 1% sodium borohydride solution for 15 minutes at 22°C and blocked with 2% bovine serum albumin for 20 minutes at 22°C. For the microtubule targeting assay, coverslips were incubated with monoclonal anti-tubulin antibodies (clone YL1/2, generously provided by J. V. Kilmartin, Laboratory of Molecular Biology, Cambridge, UK) for 30 minutes at 37°C followed by secondary donkey anti-rat antibodies conjugated to Cy5 (Jackson ImmunoResearch Laboratories, West Grove, PA) and phalloidin conjugated to Alexa Fluor 488 (Molecular Probes Invitrogen). For total internal reflection fluorescence microscopy studies, the coverslips were incubated with anti-tubulin antibodies followed by secondary goat anti-rat antibodies conjugated to Alexa Fluor 488 (Molecular Probes Invitrogen) and phalloidin conjugated to TRITC (Molecular Probes Invitrogen). For staining of adhesion markers the coverslips were incubated with mouse anti-focal adhesion kinase (FAK) antibodies (BD Biosciences, San Jose, CA) followed by donkey anti-mouse secondary antibodies conjugated to TRITC (Jackson Immunoresearch) and phalloidin conjugated to Alexa Fluor 350 (Molecular Probes Invitrogen). For quantification of CLIP and EB protein knock down, cells were treated with cold methanol (-20°C) for 10 minutes, fixed with 4% formaldehyde and permeablized with 0.1% Triton X-100. Samples were blocked with bovine serum albumin and incubated with rabbit polyclonal antibodies reactive with the head domain of CLIP170 and CLIP115 (Hoogenraad et al., 2000) or monoclonal anti-EB protein antibodies (Komarova et al., 2005). After antibody incubation, coverslips were washed in PBS and mounted onto slides using Aqua poly/mount (Polysciences, Warrington, PA).

Observations of the fixed cells for the targeting assay were performed on a Nikon Diaphot 300 inverted microscope equipped with a Plan 100×, 1.25 NA oil immersion objective. A Cy5 filter set (Chroma Technology Corporation, Brattleboro, VT) was used for observation of Cy5-labeled microtubules and a FITC filter set (Chroma Technology Corp., Brattleboro, VT) for observation of Alexa Fluor-488-phalloidin staining. 16-bit depth images were acquired with a CH350 slow scan, cooled CCD camera (Photometrics Ltd, Tucson, AZ) operated by Metamorph imaging software (Universal Imaging, Westchester, PA). Total internal reflection fluorescence microscopy was performed on an Olympus IX70 inverted microscope equipped with an Apo 100×, 1.65 NA oil immersion objective. A FITC filter set was used for observation of Alexa Fluor 488 microtubules and a Cy3 filter set (Chroma Technology Corp.) was used for observation of TRITC-conjugated phalloidin staining. Images were acquired with a 12-bit depth cooled CCD camera (Princeton Instruments Inc., Trenton, NJ) operated by Metamorph imaging software. The adhesion site marker studies were performed on a Nikon Eclipse TE200 inverted microscope equipped with a 100× oil immersion objective. A Cy3 filter set was used for observation of TRITC-labeled adhesion site markers. 16-bit images were acquired with a cooled CH250 CCD camera (Photometrics) driven by Metamorph software. All images were converted to 8-bit depth, rescaled and combined in RGB format using Adobe Photoshop (Adobe Systems, Mountain View, CA). Time-lapse imaging was performed on a Nikon Eclipse TE200 equipped with a 100× oil immersion objective. A YFP filter set was used for observation of YFP-fascin and YFP-tubulin. 12-bit image stacks were acquired with a Coolsnap HQ cooled CCD camera operated by Metamorph software. The two-band Multispec Micro-Imager system (Optical Insights, Santa Fe, NM) was used for dual channel simultaneous imaging of mRFP-actin and GFP-CLIP170 in live B16F1 cells.

Quantification of targeting events was performed on digital images of cells with fluorescently labeled microtubules and filopodia according to definitions and criteria established to aid in analysis. Cell lamellipodia had to be broad (>12 μm in width) and contain multiple (more than five) filopodia to be included in the analysis. In our previous studies, we observed frequent transition of microspikes into filopodia (Svitkina et al., 2003). In this current study, for purposes of simplicity, both microspikes (F-actin bundles that do not protrude beyond the cell margin) and longer actin bundles that protrude beyond the cell margin are called `filopodia'; however, extremely short filopodia (<1.0 μm from base to tip) were excluded from analysis. The filopodium base was defined as the point where fluorescence dropped to 50% of the peak intensity of the filopodial shaft, as determined from a linescan through the longitudinal axis of the filopodium. The filopodium base was defined as the target. A line of one-pixel width connecting filopodia bases (baseline) delimited a subpopulation of microtubules called `peripheral microtubules'. A peripheral microtubule was defined as any microtubule that intersected the baseline. The number of positive targeting events (`hits') was determined by pixel overlap. A `hit' was defined as the intersection of a peripheral microtubule with a four-pixel segment at the base of a filopodium.

The P value is equivalent to the cumulative probability of observing greater than or equal the number of positive microtubules determined from our targeting assay, assuming a purely chance targeting mechanism. Data was summed for 50 cells and the binomial distribution function in Microsoft Excel software was used to evaluate the non-randomness of targeting. The chance probability of a single microtubule hitting a filopodial base is the proportion of the baseline occupied by target. For our summed data this proportion was 0.1. The number of successes equals the number of positive events, and the number of trials equals the number of peripheral microtubules. We observed 105 positive events out of 286 peripheral microtubules. Assuming chance targeting, the probability of observing 105/286 positive events or greater is <1×10^-6=P value.

We thank Shin-ichiro Kojima for expert assistance with the molecular biology and Niels Galjart for generously provided CLIP knockout cells. We thank Kristiana Kandere for careful reading of the manuscript. This work was supported in part by grants NIH NCI U54CA119341 (G.G.B.), NIH U54GM64346 (G.G.B.), AHA 0525660Z (J.M.S.) and the Center for Drug Discovery and Chemical Biology, Northwestern University, Chicago, IL.