Coordination between membrane trafficking and actin polymerization is fundamental in cell migration, but a dynamic view of the underlying molecular mechanisms is still missing. The Rac1 GTPase controls actin polymerization at protrusions by interacting with its effector, the Wave regulatory complex (WRC). The exocyst complex, which functions in polarized exocytosis, has been involved in the regulation of cell motility. Here, we show a physical and functional connection between exocyst and WRC. Purified components of exocyst and WRC directly associate in vitro, and interactions interfaces are identified. The exocyst–WRC interaction is confirmed in cells by co-immunoprecipitation and is shown to occur independently of the Arp2/3 complex. Disruption of the exocyst–WRC interaction leads to impaired migration. By using time-lapse microscopy coupled to image correlation analysis, we visualized the trafficking of the WRC towards the front of the cell in nascent protrusions. The exocyst is necessary for WRC recruitment at the leading edge and for resulting cell edge movements. This direct link between the exocyst and WRC provides a new mechanistic insight into the spatio-temporal regulation of cell migration.
Cell motility is involved in a large variety of biological phenomena, from embryogenesis to cancer progression. Different molecules and pathways have been linked to cell migration, but an integrated and dynamic view of their interplay in the spatio-temporal space of the cell is still missing (Anitei et al., 2010; Ridley et al., 2003; Welf and Haugh, 2011).
In motile cells, the Rac1 GTPase drives the formation of the actin polymerization network at front protrusions by interacting with its effector, the Wave regulatory complex (WRC). The WRC is composed of a Wave protein [Wave1, Wave2 or Wave3, also known as WASF1, WASF2 and WASF3, respectively], Abi (Abi1, Abi2 or Abi3), Nap1 (also known as NCKAP1), Cyfip1 (also called Sra1 or Pir121) and HSPC300 (also called Brick1). These five proteins constitutively and strongly associate in the cell to form the pentameric heterocomplex (Eden et al., 2002; Gautreau et al., 2004). The WRC is intrinsically inactive (Derivery et al., 2009; Ismail et al., 2009), and at least three signals are required to fully stimulate Wave activity: phosphorylation by various kinases, including Abl and Erk proteins (Leng et al., 2005; Mendoza et al., 2011); interaction with Rac1-GTP (through the Cyfip1 subunit); and binding of phosphatidylinositol 3-phosphate phospholipids (Derivery and Gautreau, 2010; Lebensohn and Kirschner, 2009). The C-terminal verprolin homology, connector, acidic region (VCA) of the Wave subunit is sequestered according to the crystal structure of inactive WRC (Chen et al., 2010). Upon activation, the VCA domain is exposed and becomes available to bind to the Arp2/3 complex (which comprises seven subunits, two of which are ARPC2 and ARPC3) and to stimulate the nucleation of new actin filaments. Although the biochemical mechanisms of WRC activation have been largely clarified, very little is known about the spatio-temporal regulation of WRC localization.
Various studies, including those of our laboratory, have identified a migration-regulating pathway that emanates from the RalB GTPase and its downstream effector, the exocyst complex (Lim et al., 2006; Oxford et al., 2005; Rossé et al., 2006), which comprises eight subunits in yeast (Sec3, Sec5, Sec6, Sec8, Sec10, Sec15, Exo70 and Exo84). The exocyst, evolutionarily conserved from yeast to mammals, is required for polarized exocytosis – it mediates the targeting and tethering of post-Golgi secretory vesicles to specific membrane sites (He and Guo, 2009). Intracellular membrane trafficking, with its cycles of endocytosis and exocytosis, is a master organizer of signaling pathways and plays a crucial role in trafficking specific molecules to the right place at the right time (Scita and Di Fiore, 2010). With regard to cell motility, exocyst has been proposed to promote the transport and recruitment of regulatory molecules to the leading edge of migrating cells. For example, the RacGAP SH3BP1 binds to exocyst to be transported to the front of the cell, where it inactivates Rac1 (Parrini et al., 2011); similarly, the RhoGEF GEF-H1 interacts with exocyst possibly to be delivered to specific sites, where it activates RhoA (Biondini et al., 2015; Pathak et al., 2012).
In this work, by searching for new connections between the Rac1–WRC and RalB–exocyst pathways in a series of screens performed in the laboratory, we discovered a much unexpected link – the WRC and exocyst directly interact. Here, we report the biochemical characterization of this interaction and its functional relevance for cell motility. This finding provides a new molecular basis underlying the spatio-temporal coordination of membrane trafficking and actin polymerization in motile cells.
Exocyst and WRC directly interact
A series of yeast two-hybrid screens aimed at identifying human proteins interacting with the human exocyst indicated that exocyst and WRC might associate through two points of contact (Fig. 1A). In particular, the Abi1 and Abi2 subunits of the WRC were recovered from yeast two-hybrid screens with full-length human EXOC7 variant 1 (the homolog of Saccharomyces cerevisiae Exo70). A second WRC subunit, Cyfip1, was found in screens performed with human EXOC3 (the homolog of S. cerevisiae Sec6; fragment comprising residues 297–745). Yeast nomenclature has been used for human exocyst proteins herein. The deduced minimal Exo70-interacting domain of Abi1 and Abi2 corresponded to residues 37–188 (total length of Abi proteins is 329–513 residues, depending on the isoform). By contrast, the minimal Sec6-interacting domain of human Cyfip1 (1253 residues) comprised the residues 326–508, which overlap with the Cyfip1 region that binds to Rac1-GTP (Kobayashi et al., 1998) (Fig. 1B).
To validate these interactions biochemically, we performed in vitro binding assays with purified proteins. Exo70 and Sec6 were purified from E. coli as GST-fusion products, whereas Abi1 was purified from insect cells. By using GST pulldown assays, we found that GST–Exo70 efficiently bound to purified Abi1 to extent similar to EPS8, a well-characterized Abi1-interacting protein (Disanza et al., 2004), demonstrating the direct interaction between Exo70 and Abi1 (Fig. 1C). Such a simple strategy could not be applied to the Sec6–Cyfip1 interaction because isolated Cyfip1 is unstable. Because Cyfip1 folding can be stabilized by the association to Nap1, we co-overexpressed in HEK293 cells Flag-tagged Nap1 together with Myc-tagged Cyfip1 and co-isolated a bi-molecular Nap1–Cyfip1 complex by using anti-Flag immunoprecipitation, as previously reported (Innocenti et al., 2004). The bead-bound Nap1–Cyfip1 complex was incubated with purified GST–Sec6 and GST–Exo70 (Fig. 1D) – Sec6 (lane 8) but not Exo70 (lane 7) associated with Nap1–Cyfip1; neither Exo70 nor Sec6 associated with Nap1 alone (lanes 5 and 6) nor with a control unrelated Flag–MEK1 (human MAP2K1) fusion (lanes 3 and 4). These results indicate that Cyfip1, but not Nap1, interacts with Sec6 but not with Exo70, confirming the two-hybrid analysis findings. Consistently, the whole pentameric WRC binds in vitro to both GST–Sec6 (through Cyfip1) and GST–Exo70 (through Abi1) but not to GST alone (Fig. S1A and B).
Our two-hybrid data indicate that the region of Cyfip1that binds to Sec6 overlaps with the site in Cyfip1 required to bind to active Rac1 (Kobayashi et al., 1998), suggesting that active Rac1 might compete with exocyst for binding to WRC. Competition experiments using purified proteins, however, did not support this possibility. Indeed, soluble GST–Sec6 bound to immobilized WRC even in the presence of a large molar excess of soluble active Rac1 (Q61L mutant), indicating that active Rac1 cannot prevent the association of WRC with Sec6, at least in vitro, and suggesting the possible existence of a super-complex comprising exocyst–WRC–Rac1-GTP in vivo (Fig. 1E).
Exocyst and WRC associate in cells
To study WRC and exocyst complexes in cells, we used HEK293 cells stably expressing tagged human subunits (Flag–HA–Cyfip1 and Flag–HA–Abi1 for WRC; V5–Sec6 and V5–Sec5 for exocyst). The level of expression of exogenous proteins was comparable to that of endogenous ones. We confirmed the previously reported observation (Derivery et al., 2009) that tagged Cyfip1 and Abi1 replace the endogenous subunits in the WRC (Fig. S1A). Moreover, we observed that, in the presence of V5–Sec6, endogenous Sec6 is no longer detectable in the whole lysates, suggesting that a similar molecular replacement mechanism occurs for the exocyst complex (Fig. 2B).
When immunoprecipitations of Flag were performed on lysates from cell lines expressing Flag–HA–Cyfip1, the whole WRC was retrieved as expected – bands representing Cyfip1, Nap1, Wave2 and Abi1 appeared in an approximately stoichiometric ratio, according to the quantification on a Ponceau-Red-stained gel (note that Brick1 runs off standard gels because of its small size). Importantly, endogenous Exo70 and Sec8 co-immunoprecipitated with the WRC from Flag–HA–Cyfip1-expressing cells, but not from control cells expressing Flag–HA tag only (Fig. 2A). However, in the immunoprecipitates, we could not detect Sec6, the direct interactor of Cyfip1. This is likely to be due to the weak sensitivity of the available antibodies against Sec6. Endogenous Exo70, the direct interactor of Abi1, was also found to be associated with the WRC isolated from Flag–HA–Abi1-expressing cells (Fig. S1C).
In parallel, immunoprecipitations of V5 were performed on lysates from cell lines expressing V5–Sec6. The stoichiometry of the retrieved exocyst complex was not equimolar, the Sec6 subunit being in excess with respect to other components. This was not surprising considering the dynamic nature of this complex. All exocyst subunits for which antibodies were available (Sec5, Sec6, Sec8, Exo70, Exo84) were present in the immunoprecipitates. Endogenous Cyfip1 protein specifically co-immunoprecipitated with this Sec6-enriched exocyst complex (Fig. 2B), which is consistent with the existence of a Sec6–Cyfip1 interaction in cells. In addition, we also specifically detected Wave2 and Nap1 in the exocyst precipitates from both V5–Sec6- and V5–Sec5-expressing cells (Fig. 2C). In conclusion, no matter which subunit of WRC (Cyfip1 or Abi1) or exocyst (Sec6 or Sec5) was precipitated, components of the other complex were detected (Sec8 and Exo70 precipitated with WRC subunits; and Cyfip1, Wave2 and Nap1 precipitated with exocyst components), suggesting that assembly occurs between the whole complexes and not simply between isolated subunits.
The Arp2/3 complex is a shared interacting complex of both WRC and Exo70. Indeed, W. Guo and colleagues have shown that Exo70 directly binds to the ARPC1 subunit of the Arp2/3 complex, and promotes actin branching by acting as a kinetic activator of Arp2/3 (Liu et al., 2012; Zuo et al., 2006). To directly assess a possible involvement of Arp2/3 as bridge in mediating the WRC–exocyst interaction, we treated our cells with small interfering (si)RNAs targeting the crucial subunit ARPC2, a condition known to induce degradation of the entire Arp2/3 complex (Abella et al., 2016). We verified in particular the depletion of the subunits ARPC1A and ARPC1B, which interact with Exo70. Under these conditions, we still observed co-precipitation of Exo70 upon precipitation of the WRC, providing evidence against the notion that Arp2/3 mediates the interaction between WRC and exocyst (Fig. 2D).
We found that our cells mainly expressed the Exo70 isoform 2 [also known as the mesenchymal (M) variant, denoted Exo70-2M] (Fig. S1D), which has been reported to bind to Arp2/3 (Lu et al., 2013). On the contrary, the Exo70 isoform 5 [also known as the epithelial variant (E), denoted Exo70-5E], carrying a 23-amino-acid insertion conferred by alternative splicing, does not bind to Arp2/3 (Lu et al., 2013). We expressed GFP-fused Exo70-5E and Exo70-2M in the HEK293 Flag–HA–Abi1-expressing cell line. Both isoforms were found to be associated with the WRC that could by immunoprecipitated with an antibody against Flag (Fig. S1E). The observation that Exo70-5E, which cannot bind to Arp2/3, is still able to bind to WRC, indicates that the interaction between Exo70 and WRC occurs independently of the interaction between Exo70 and Arp2/3.
We reasoned that the exocyst and WRC might interact locally and/or transiently, possibly during cell migration. Hence, we focused our investigation on motile cells, using HEK-HT cells (which ectopically express hTERT and large and small T-antigen) as motility model (Hahn et al., 1999). Randomly moving HEK-HT cells were fixed and co-stained with antibodies against Wave2 and against Exo70 or Sec6. All three proteins were found at cell protrusions in agreement with previous reports (Nozumi et al., 2003; Rossé et al., 2006; Zuo et al., 2006), with extensive apparent overlap, as shown by confocal microscopy analysis (Fig. 3). However, their patterns were different – Wave2 was clearly located at the very edge of lamellipodia, with a sharp and continuous signal; the front-edge staining of Sec6 and Exo70 appeared, instead, blurred and discontinuous. Finally, almost 50% of the protrusions positive for Wave2 staining were devoid of Exo70 or Sec6 (n=35 protrusions from 21 cells). In Fig. 3, we show representative fluorescence intensity plots of both exocyst and WRC overlapping staining (region a) and WRC-only staining (region b).
These data suggest that the active WRC is not bound in a compulsory manner to the exocyst at the edge. Because Rac1 activates the WRC by interacting with Cyfip1, it is possible that WRC is retained by Rac1, whereas the exocyst recycles in the cytosol.
Mapping of the Exo70–Abi interaction surface
To identify the Exo70–Abi interaction surface, we co-overexpressed wild-type and deletion mutants of GFP-fused Abi1 and Exo70–HA constructs, and we performed immunoprecipitations for GFP.
For Abi proteins, we found that the full-length, but not a mutant lacking the N-terminal domain of Abi1 (residues 1–145), co-precipitated with Exo70–HA. The N-terminal domain of Abi2, which is nearly identical to that of Abi1, was sufficient for a very weak interaction (Fig. 4A). This indicates that the N-terminal domain of Abi proteins is necessary and even sufficient to some extent to drive the interaction between Exo70 and Abi, consistent with the two-hybrid data.
To further investigate how the Abi N-terminal could interact with Exo70 in the context of the WRC assembly, the conservation of residues accessible at the surface of Abi1 and Abi2 in WRC was analyzed using the Rate4Site algorithm (http://www.tau.ac.il/~itaymay/cp/rate4site.html) applied on a multiple sequence alignment of 150 homologs of Abi1 and Abi2 in animals (Pupko et al., 2002). A conserved surface patch could be delineated between the helices of Abi1 spanning the regions comprising residues 18–36 and 47–65 (Fig. 4B). At the center of this patch, the residue Q56 is exposed and nearly invariant. We generated the mutation Q56A together with two other mutations, Y32E and Y63E, corresponding to exposed and conserved aromatic residues at the edge of the conserved patch. Only the Q56A mutation substantially decreased the interaction with Exo70 (Fig. 4C), showing that this exposed position plays a key role in bridging WRC and Exo70.
For Exo70, the N-terminal Exo70 moiety (residues 1–414), but not the C-terminal Exo70 moiety (residues 415–C-terminus), binds to full-length Abi1. Exo70 has an N-terminal coiled-coil motif as predicted using the PCOILS program (https://toolkit.tuebingen.mpg.de/pcoils) (Lupas et al., 1991). The deletion of this N-terminal coiled-coil motif (residues 1–84; Exo70-Δ84) abolished the binding to Abi1 (Fig. 4D). A very similar sequence (residues 1–75) has been reported to be necessary for Exo70 dimerization (Zhao et al., 2013).
Exocyst–WRC association contributes to cell migration
We used the Exo70-Δ84 mutant, which is deficient in binding to Abi1, to investigate the functional consequence of impairing exocyst–WRC association during motility. In HEK-HT cells, as expected (Zuo et al., 2006), the depletion of Exo70 inhibited the speed of cell motility. This phenotype was rescued by re-expression of wild-type Exo70 but not by the re-expression of the Exo70-Δ84 mutant (Fig. 5A and B). Thus, the N-terminal region of Exo70 is necessary for efficient cell motility, in agreement with a previous report that used an Exo70-Δ75 mutant (Zhao et al., 2013). Importantly, the depletion of the N-terminus of Exo70 (residues 1–75) was not shown to affect exocyst assembly nor its function in exocytosis (Zhao et al., 2013). Therefore, the motility defect of cells expressing N-terminal truncated Exo70 mutants (Δ84 or Δ75) is not due to a general loss-of-function of the entire exocyst complex. Our findings are consistent with the interpretation that the depletion of the Exo70 N-terminal region inhibits cell migration at least in part because it impairs the exocyst–WRC association, supporting a role for exocyst–WRC association during cell migration.
Identification of a forward flow of WRC
We reasoned that the requirement of an exocyst–WRC association for efficient migration might underlie a function of exocyst in the dynamics of WRC localization. To directly visualize the spatio-temporal dynamics of the WRC in motile HEK-HT cells, we exogenously expressed a GFP–Abi1 fusion. Importantly, the expression level of GFP–Abi1 was kept as low as possible to favor its incorporation into the endogenous WRC. With such a precaution, by confocal spinning-disc fast-acquisition imaging, we clearly observed a sharp recruitment of GFP–Abi1 at the edge of nascent and expanding protrusions (Fig. 6A; Movie 1), as previously reported (Stradal et al., 2001).
In regions just behind the active edges, WRC live localization appeared like a dense meshwork of fluorescent highly dynamic cytosolic dots. As originally reported using a single-molecule imaging approach (Millius et al., 2012), we also observed rare events of WRC retrograde motion, which we manually tracked to find a median speed of 0.161 µm/s (s.d.=0.036, n=13 dots from five cells); this is consistent with the notion that a portion of the WRC population is recycled through incorporation into the actin network, undergoing retrograde flow (Millius et al., 2012). However, most of the movements of these Abi-positive dots at protrusions were not trackable because of their high density and their rapid movements. Therefore, we took advantage of the spatio-temporal image correlation spectroscopy technique (STICS) (Hebert et al., 2005). STICS allows measurement of both diffusion coefficients and velocity vectors (magnitude and direction), or ‘molecular flow’, for fluorescently labeled proteins in living cells through space–time correlation functions.
We applied STICS analysis to the front portions of motile cells expressing GFP–Abi1. Fig. 6B shows an example of estimation, both in space and in time, of Abi1 flow by STICS analysis in a representative cell. The displacement of the peak of the spatial autocorrelation function (ACF) indicates both the direction of flow and its magnitude (Hebert et al., 2005).
The velocity of the molecular WRC flow was in the range 0.001–0.09 µm/s, with a median of 0.0215 µm/s (s.d.=0.023, n=82 measurements from 21 regions of interest at various time intervals, from eight independent cells). These measured flows were not simply due to cytosolic motions because the same analysis on control cells (expressing GFP) only very rarely could detect a molecular flow that was always very slow (median 0.005 µm/s, s.d.=0.005, n=22 measurements from six regions of interest from three control GFP cells). Globally, the WRC flows did not show a preferential directionality with respect to the cell front. We therefore divided the protrusions into three categories according to their status: nascent, active and ending (see Materials and Methods for technical definitions). This categorization clearly revealed that the nascent protrusions present a strong dominance of consistent anterograde flow of WRC (Fig. 6C; Fig. S2A and B).
In parallel, we attempted to apply a similar approach to cells expressing GFP–Sec6, with the aim of correlating WRC and exocyst translocation dynamics. Disappointingly, neither by correlation methods nor by particle tracking, we could not obtain any evidence of Sec6 directed flows, instead, we found a diffusion-like motion (Fig. S3A, B and C; Movie 3). These results indicate that exocyst trafficking, even if presumably polarized toward the front, remains very difficult to visualize, possibly because of the continuous fast recycling of this complex (Rivera-Molina and Toomre, 2013) or because the exogenous GFP–Sec6 cannot fully recapitulate the behavior of the endogenous Sec6.
In conclusion, by using this novel correlation-based image analysis approach, we could show that in nascent protrusions, the majority of WRC molecules travel toward the front edge, where they are subsequently recruited.
Correlation between WRC recruitment and edge velocity at protrusion sites
We quantified the dynamic recruitment of GFP–Abi1 at protrusive sites by precisely measuring in time (images every 2 s) and space (100× magnification) the intensity of the GFP fluorescence at the cell periphery. In parallel, we also measured the velocity of the edge movements. To do so, we implemented a freely available ImageJ plugin (see Materials and Methods). This plugin allows computation of the fluorescence intensity at the cell edge and the edge velocity, for each time point and for each position along the edge. An overview of the results is given by kymograph representations, as shown in Fig. 7A, for a representative cell region (Movie 2) – GFP–Abi1 (i.e. WRC) displays a dynamic spatio-temporal profile of recruitment at protrusion edges, which mirrors the spatio-temporal profile of edge velocity.
To mathematically confirm these observations, we computationally calculated the spatio-temporal correlation between WRC recruitment and edge velocities for several protrusive regions, following a previously reported method (Machacek et al., 2009). Notably, this analysis showed that the maximum correlation values were narrowly distributed around a time lag of 0 s (Fig. 7B). The computed averaged time delay between edge advancement and WRC recruitment (averaged time lag=−1.6 s±0.66 s, n=10 protrusions) was below our acquisition time interval (2 s). This finding indicates that WRC recruitment is synchronous with edge movement, at least within the limit of our temporal resolution. In other words, as soon as WRC is recruited, the edge moves forward and, as soon as WRC leaves, the edge stops moving or starts retracting. This striking spatio-temporal correlation underscores a possible cause–effect relationship between WRC translocation and protrusion.
Exocyst is required for optimal WRC recruitment and edge velocity
We questioned the necessity of the Ral–exocyst pathway for optimal WRC recruitment by staining the endogenous Wave2 subunit of HEK-HT cells that had been fixed during random motility. The depletion of exocyst subunits (Sec6, Exo70 or Sec5), as well as of Ral proteins (RalA and RalB), reduced the fraction of the cell perimeter that was positive for Wave2, without substantially perturbing total cell perimeter (Fig. 8A; see Fig. S4A for validation of depletions). The expression of constitutively active Rac1V12 has been previously reported to increase WRC recruitment at the cell membrane periphery (Steffen et al., 2004), probably because of a retention effect after WRC translocation. We reproduced this finding by overexpressing RFP–Rac1V12 in our cell system (compare siControl in Fig. 8A versus B). In addition, we found that, also in Rac1V12-expressing cells, the depletion of exocyst subunits or of Ral proteins reduced the staining of Wave2 at the cell perimeter (Fig. 8B; Fig. S4A), indicating that both exocyst and active Rac1 are required for maximal WRC recruitment.
The observation that both RalA and RalB depletion affects WRC recruitment was surprising because RalB, but not RalA, has been reported to be necessary for motility in various cell models (Lim et al., 2006; Oxford et al., 2005; Rossé et al., 2006). When we assessed the effect of Ral silencing on the motility of HEK-HT cells, we observed an inhibition upon RalA depletion, which was milder than that upon RalB depletion (Fig. S4B). This suggests that RalA could, indeed, have a role in cell motility in HEK-HT cells.
In addition, we performed experiments in live cells using Cherry–Abi1 as a quantifiable sensor of WRC recruitment. In selected front regions, as in Fig. 7A, we computed Cherry–Abi1 fluorescence at the cell edge, and we measured the fluctuation of WRC recruitment by calculating the coefficient of variation of Cherry–Abi1 fluorescence over the time and over the edge length. In cells depleted of Sec6 or Exo70, there was a significant decrease in the fluctuation of WRC recruitment with respect to control cells (Fig. 8C; Movie 4), indicating that exocyst is necessary for the normal dynamics of WRC localization at leading edges. Consistent with the correlation described above between WRC recruitment and edge velocity, this defect of WRC recruitment dynamics was associated with a defect in edge velocity – the edges of Sec6- or Exo70-depleted cells protruded and retracted more slowly than control cells (Fig. 8D; Movie 4). Interestingly, both positive (protrusions) and negative velocities (retractions) were decreased in exocyst-depleted cells, suggesting that back and forward edge movements are intimately connected. Finally, as expected (Rossé et al., 2006), silencing of exocyst subunits inhibited cell velocity in a wound healing assay (Fig. S4B).
These results together strongly support the notion that exocyst promotes WRC recruitment at the leading edge, which in turn causes edge advancement, initiating a protrusion event and stimulating cell motility.
In this work, we show the existence of a physical association between exocyst, an important complex in exocytosis, and WRC, a crucial module in actin polymerization machinery, and we provide evidence for a functional role of this association in the recruitment of WRC to specific plasma membrane sites where protrusions are formed in migrating cells.
How is WRC transported to the cell periphery and recruited to the leading edge of lamellipodia? This is a very crucial question in order to understand cell motility. Kinesin-mediated transport along microtubules has been proposed to bring WRC to the cell periphery in motile cells (Takahashi and Suzuki, 2008, 2009). Very interestingly, the exocyst has also been proposed to use microtubules tracks to target secretory vesicles. Exocyst has been shown to associate and colocalize with microtubules (Vega and Hsu, 2001; Xu et al., 2005). In neuronal cells, this association promotes neurite outgrowth (Vega and Hsu, 2001). Therefore, both WRC and exocyst might exploit microtubules for the long-distance travel to protrusions. Our work suggests that the direct interaction between WRC and exocyst might play a role in the short-distance travel of WRC inside the protrusions, resulting in its recruitment to the leading edge.
Other non-exclusive molecular mechanisms have been proposed for the role of exocyst in protrusion initiation and motility. In particular, two properties of Exo70 link this exocyst subunit to the migration machinery. Firstly, Exo70 directly binds to and stimulates the actin-nucleating Arp2/3 complex (Liu et al., 2012; Zuo et al., 2006); however, N-terminal truncation of Exo70 impairs cell migration despite the fact that Arp2/3 binding and stimulation are normal (Zhao et al., 2013), indicating that Exo70 can regulate cell migration also by a mechanism independent of direct Arp2/3 stimulation. Secondly, Exo70 can generate membrane curvatures in vitro and in vivo through an oligomerization-based mechanism. This Exo70 function does require the Exo70 N-terminal region (Zhao et al., 2013). Therefore, the Exo70 N-terminal coiled-coil region can mediate Exo70–Abi1 interaction, promoting exocyst–WRC association, as well as Exo70–Exo70 interactions, generating membrane curvature.
The X-ray structure of the whole exocyst complex has not been solved yet. However, the observation that multiple exocyst subunits can bind to the same partner is informative – Sec5 and Exo84 to Ral proteins (Moskalenko et al., 2003), Exo70 and Sec6 to WRC (this work), Exo84 and Sec8 to SH3BP1 (Parrini et al., 2011), Exo84 and Sec3 to WASH (Monteiro et al., 2013). This suggests that the exocyst subunits are intricately packed together, exposing composite surfaces for interaction with partners, as supported by the elongated exocyst architecture recently observed by using negative-stain electron microscopy (Heider et al., 2016).
We showed that the N-terminal coiled-coil motif of Exo70 (residues 1–84) and the N-terminal of Abi proteins (residues 1–146) are required for the Exo70–Abi interaction. Noteworthy, in the WRC, the N-terminal region of Abi2 was found to fold as a heteromeric coiled-coil structure with Brick1 and Wave1 (Chen et al., 2010). Interestingly, in the context of SNARE complexes, Abi1 has been reported to bind to syntaxin-1 and SNAP-25 through a region spanning N-terminal residues 54–108 (Echarri et al., 2004). The latter two proteins are known to interact with their partners through coiled-coil bundles (Sutton et al., 1998). Altogether, these elements suggest that the N-terminal regions of Abi1 and Abi2 are prone to form triple helix coiled-coils with a versatile set of binding partners, and point to a possible similar mechanism for the interaction with Exo70. The roles of Exo70 dimerization, its incorporation into the exocyst complex and its interaction with the WRC remain to be clarified.
MATERIALS AND METHODS
Yeast two-hybrid methods
The coding sequence for full-length human EXOC7 variant 1 (homolog of S. cerevisiae Exo70) (GenBank accession: NM_001013839) was cloned into pLex12 as a C-terminal fusion to LexA. The coding sequence for residues 297–745 of human EXOC3 (homolog of S. cerevisiae Sec6) (GenBank accession: NM_007277) was cloned into pB27 as a C-terminal fusion to LexA. The constructs were used as a bait to screen at saturation a highly complex randomly primed human placenta cDNA library constructed into pP6. About 100 million clones (seven times the complexity of the library) were screened using a mating approach with Y187 (matα) and L40ΔGal4 (mata) yeast strains, as previously described (Formstecher et al., 2005; Fromont-Racine et al., 1997). 380 colonies in the Exo70 screen and 190 colonies in the Sec6 screen were selected on a medium lacking tryptophan, leucine and histidine. The prey fragments of the positive clones were amplified with PCR, sequenced at their 5′ and 3′ junctions and identified in GenBank (NCBI).
Cell culture, transfections, plasmids and siRNAs
Cells were grown in Dulbecco's modified Eagle's medium supplemented with 2 mM glutamine, penicillin, streptomycin, 10% fetal bovine serum. See Table S1 for detailed information and selection antibiotics. All cell lines were routinely verified to be free of mycoplasma by using a PCR detection method (Venor®GeM Classic, Minerva Biolabs).
In vitro interaction assays
GST–Sec6 and GST–Exo70 were produced in Rosetta2 or BL21 E. coli strains by addition of 0.5 mM IPTG at 20–25°C for 4 h. Only a fraction of the expressed GST fusion was soluble, leading to poor yields (about 50 µg per 1 liter of culture). Abi1, the bi-molecular Nap1–Cyfip1 complex and the whole pentameric WRC were purified as previously reported (Innocenti, 2004; Derivery et al., 2009). The active Rac1 protein (Q61L mutant) was purified from E. coli (Self and Hall, 1995).
For the interaction assays shown in Fig. 1C,D and Fig. 1B, protein-loaded beads and purified proteins were incubated for 2 h at 4°C in a 300-µl final volume of interaction buffer A (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1% Triton X-100, 5% glycerol, freshly supplemented with protease inhibitor cocktail and 1 mM DTT). The amount of the input soluble proteins (Abi1, GST–Sec6, GST–Exo70) was 10–30 µg.
The competition assay in Fig. 1E was performed for 2 h at 4°C in a 300-µl final volume of interaction buffer B (20 mM Tris-HCl pH 8, 150 mM NaCl, 10% glycerol, 5 mM MgCl2, freshly supplemented with protease inhibitor cocktail and 1 mM DTT).
For the immunoprecipitations, protein lysates from two 15-cm dishes were prepared in 1 ml per dish of lysis buffer C (50 mM Hepes, 150 mM NaCl, 1% NP-40, 0.1% sodium dodecyl sulfate, 0.5% sodium deoxycholate, 5 mM EDTA, pH 7.7, freshly supplemented with protease inhibitor cocktail and 1 mM DTT). For immunoprecipitation of Flag, lysates were incubated for 2 h at 4°C with 25 µl of agarose beads or magnetic beads coupled to FlagM2 antibody (Sigma-Aldrich), then washed five times with 1 ml of lysis buffer C and finally boiled in gel-loading buffer. For immunoprecipitation of V5, lysates were incubated (i) for 2 h at 4°C with 7 µl of agarose beads coupled to anti-V5 antibody (Sigma-Aldrich) or (ii) for 1 h at 4°C with 10 µg of anti-V5 antibody (Invitrogen), followed by 1 h at 4°C with 20 µl of magnetic beads coupled to G protein (Invitrogen). Beads were washed three times with high-salt buffer (lysis buffer with 500 mM NaCl) and three times in low-salt buffer (lysis buffer lacking NaCl).
For immunoprecipitation with overexpressed proteins (Fig. 4), lysates from one 6-cm dish were prepared in 0.6 ml lysis buffer D (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1% Triton X-100, 1 mM EDTA, freshly supplemented with protease inhibitor cocktail and 1 mM DTT). Lysates were incubated for 1 h at 4°C with magnetic beads coupled to anti-GFP antibody (ChromoTek). Beads were washed three times with the same buffer.
Cells were fixed with 4% paraformaldehyde for 15 min and permeabilized for 5 min with 0.1% Triton X-100 or 0.5% NP40. See Table S4 for antibodies. In order to quantify the recruitment of Wave2 at the cell edges of random motile cells (Fig. 6A and B), for each cell, we manually draw and measured, by using ImageJ software, the total cell perimeter (a) and the perimeter portion positive for Wave2 staining (b). The ‘Wave-positive perimeter’ percentage was calculated as (b/a)×100.
See Table S4 for antibodies. Detection was achieved with the ECL chemiluminescence method (PerkinElmer, #NEL104001EA) using horseradish peroxidase (HRP)-conjugated secondary antibodies (Jackson ImmunoResearch) or with the LICOR Odyssey Infrared Imaging System (LI-COR Biosciences) using IRDye-conjugated secondary antibodies.
Fast-acquisition spinning-disc microscopy
Cells were imaged on a Nikon Ti inverted microscope equipped with an EMCCD camera, a spinning disc (Yokogawa CSU-X1) with four-laser line, a heated chamber and CO2 controller (Life Imaging Services), under control of the MetaMorph software (Universal Imaging). Images were acquired every 1 or 2 s, using a 100× NA1.4 oil immersion objective.
Spatio-temporal image correlation spectroscopy
Protrusions were searched for visually in the spinning-disc movies. STICS measurements were performed on a region of interest (ROI) of 32×32 pixels (5.12×5.12 µm), with a ROI center at about 3–5 µm and a ROI edge at ∼1–2 µm from the closest cell edge, in order to avoid a border effect. If obvious vesicles were visible in the ROI, the measurements were discarded. We developed and used the plugin MICS Toolkit (downloaded from http://minilien.curie.fr/dr8ep8), based on previous work (Brown et al., 2006; Kolin and Wiseman, 2007). First, image correlation spectroscopy (ICS) was performed on a set of three images in order to check for the point spread function (PSF) of the experiment. Then, STICS analysis was performed on the selected ROI on time windows of 20 frames (20–40 s depending on the microscope acquisition setting). Gaussian fitting of the spatiotemporal correlation function was visually checked, and only the lag times for which it appeared correct were considered for the linear fitting of the displacement of the peak. The ROI was considered as having a flow only if the displacement of the peak against lag times was monotonic in both x and y. An average velocity for the selected time period and a direction were then obtained. The velocity values were coherent with temporal image correlation spectroscopy (TICS) measurements of a population undergoing both diffusion and flow. We used homemade in silico simulations, using parameters (density, diffusion, flow) similar to those observed in cells, to validate the robustness of the method (not shown).
A time widow of a specific ROI was classified as ‘nascent protrusion’ when the last two to three frames (out of 20) showed the front edge starting to move forward. It was classified as an ‘active protrusion’ when the front edge was actively moving during the entire sequence. It was classified as ‘ending protrusion’ when retraction was clearly observed during the temporal sequence. When possible, the same ROI was analyzed in the various protrusion states.
Cell-edge tracking, recruitment measurement and cross-correlation analysis
Cells were cropped around the protrusions. The edge segmentation over time was achieved by creating and using an ImageJ threshold function, and a time-lapse binary mask, which was visually checked and manually corrected when needed. Starting from the original image and from the binary mask, we then used a custom-made ImageJ plugin, named Recruitment Edge Dynamics (which can be downloaded from http://minilien.curie.fr/pnj9m7), which tracks, over the time, uniformly sampled points (windows) to generate two maps. The first map represents the protein recruitment – i.e. the maximum fluorescent intensity at the edge (width of 0.8 µm) for each window (in y) and for each timeframe processed (in x). The second map represents the speed of the edge for the same times and windows; positive speed indicates a protrusion, speed equal to zero indicates a static edge, whereas negative speed indicates a retraction. For cell-edge tracking, an approximate center of gravity of the cell was manually defined (outside of the image because cells had been cropped); a line passing by this center of gravity and by the sampled point at t was used to define the point position at t+1 as its intersection with the edge at t+1.
To compute the temporal cross correlation between protein recruitment and edge speed, we used a custom-made MATLAB program [which can be downloaded from http://xfer.curie.fr/get/Zn3LkZP0wDG/Correlation.zip, for Windows 64bits only, need MRC installer version 8.1 (2013 A) available freely from the MathWorks website, source code at http://xfer.curie.fr/get/Zn3LkZP0wDG/Correlation.zip]. Similar to the method of Machacek et al. (2009), we averaged by spline fitting the correlation of edge speed and recruitment over time for each protrusion, and we considered the correlation as significant if it passed the two following criteria: (1) the averaged maximum value of the cross-correlation coefficient was superior to the 90% confidence level of the correlation coefficient, depending on the number of sampled points; (2) the confidence interval of the spline fitting, computed by bootstrapping 2000 samples from the residuals, did not include 0. Correlation curves for individual protrusions were then merged to obtain the averaged maximum of correlation, corresponding to the time lag, and to calculate a confidence interval for this value.
We greatly thank the staff of the PICT-IBiSA imaging facility for their excellent assistance, Hybrigenics staff for yeast two-hybrid analysis and Nathalie Brandon for technical help. Some experiments were performed on the PICT-IBiSA (Institut Curie, Paris), member of the French National Research Infrastructure France-BioImaging (ANR-10-INSB-04).
M.B., A.S.-D., G.Z., M.A., M.C.P. performed the biological experiments. P.P.-G., F.W., M.C.P. developed image analysis codes and analyzed the data. E.F. coordinated 2-hybrid screens. M.H., A.G. generated cell lines. J.Y., R.G. performed structural modeling. M.C.P., G.S., J.C., A.G. designed and supervised the experiments. M.C.P. wrote the manuscript. A.S.-D., P.P.-G. and G.Z. made equal contributions to this work, as did G.S., J.C. and M.C.P.
This work was supported by Association pour la Recherche sur le Cancer [grant numbers SFI20121205710 to M.C.P. and SFI20111203931 to J.C.]; Ligue Contre le Cancer [grant number RS14/75-54 to M.C.P.]; Association Christelle Bouillot (to J.C.); GenHomme Network [grant number 02490-6088, to Hybrigenics and Institut Curie); Associazione Italiana per la Ricerca sul Cancro [grant number IG 14104 to G.S.); the European Research Council [grant number 268836 to G.S.]; the Association for International Cancer Research [grant number 14-0335 to G.S.]; and the Fondazione Cariplo (to G.S.). M.B. and A.S.-D. were recipients of fellowships from Association pour la Recherche sur le Cancer, Ligue Contre le Cancer, European Molecular Biology Organization and Vinci program of the Université Franco-Italienne.
The authors declare no competing or financial interests.