Lamellipodin promotes actin assembly by clustering Ena/VASP proteins and tethering them to actin filaments

Enabled/Vasodilator (Ena/VASP) proteins promote actin filament assembly at multiple locations, including: leading edge membranes, focal adhesions, and the surface of intracellular pathogens. One important Ena/VASP regulator is the mig-10/Lamellipodin/RIAM family of adaptors that promote lamellipod formation in fibroblasts and drive neurite outgrowth and axon guidance in neurons. To better understand how MRL proteins promote actin network formation we studied the interactions between Lamellipodin (Lpd), actin, and VASP, both in vivo and in vitro. We find that Lpd binds directly to actin filaments and that this interaction regulates its subcellular localization and enhances its effect on VASP polymerase activity. We propose that Lpd delivers Ena/VASP proteins to growing barbed ends and increases their polymerase activity by tethering them to filaments. This interaction represents one more pathway by which growing actin filaments produce positive feedback to control localization and activity of proteins that regulate their assembly. DOI: http://dx.doi.org/10.7554/eLife.06585.001


Introduction
Eukaryotic cells assemble networks of actin filaments adjacent to the plasma membrane to carry out many fundamental processes, including: maintenance of cell morphology; amoeboid locomotion; and cell-cell interaction. The architecture and function of these actin networks is specified in part by the properties of membrane-associated regulatory molecules Lacayo et al., 2007). For example, new filaments are created by nucleation factors recruited to the plasma membrane by Rhofamily G-proteins, while fast-growing barbed ends of actin filaments near the membrane interact with regulatory factors that alter the rate and duration of filament elongation. Some factors, including formin-family proteins (Romero et al., 2004;Kovar et al., 2006) and Enabled/Vasodilator (Ena/VASP stimulated phosphoprotein) proteins (Barzik et al., 2005;Breitsprecher et al., 2008;Hansen and Mullins, 2010;Breitsprecher et al., 2011), accelerate filament elongation, while others, such as capping protein, terminate filament elongation (Dinubile et al., 1995;Kuhn and Pollard, 2007). Local changes in the rates of filament elongation and capping can alter the architecture and function of the actin network and thus tip the balance between membrane protrusion and retraction Applewhite et al., 2007).
One important group of actin regulatory proteins is the Ena/VASP family: weakly processive actin polymerases that accelerate filament elongation and slow filament capping (Barzik et al., 2005;Breitsprecher et al., 2008;Hansen and Mullins, 2010;Breitsprecher et al., 2011). Recruitment of Ena/VASP proteins to the plasma membrane often promotes outgrowth of thin, finger-like filopodia and sometimes promotes rapid advance of broad lamellipodial sheets (Lanier et al., 1999;Rottner et al., 1999). Little is known about the molecular mechanisms that determine localization and activity of Ena/VASP proteins in vivo, but previous work suggests that formin-family protein Diaphanous (Grosse et al., 2003;Schirenbeck et al., 2006;Barzik et al., 2014;Bilancia et al., 2014) and the WAVE regulatory complex Chen et al., 2014), relies on a related, but lower affinity, EVH1-binding motif: LPPPPP.
The MRL proteins colocalize with Ena/VASP at the plasma membrane of many different cell types in many diverse organisms, both vertebrates and invertebrates . In the nematode, Caenorhabditis elegans, the mig-10 protein promotes neurite outgrowth and proper axon guidance (Manser et al., 1997;Chang et al., 2006). Similarly, in mammals the protein Lamellipodin (Lpd) helps determine morphology of neurons and promotes growth of lamellipodial protrusions in a variety of non-neuronal cells (Michael et al., 2010;Pinheiro et al., 2011;Yoshinaga et al., 2012;Law et al., 2013). Several MRL proteins, including Lpd, contain a tandem Ras-Association and Pleckstrin Homology (RA-PH) domain which, together with an adjacent coiled-coil region, causes the proteins to form weak homo-dimers in solution (Chang et al., 2012). The RA-PH domain is structurally homologous to growth factor receptor-binding proteins Grb7, Grb10, and Grb14 (Depetris et al., 2009) but, unlike the Grb proteins, no direct interaction between small GTPases (e.g., Ras or Rho family) and MRL proteins has been reported. In vitro, the MRL pleckstrin homology (PH) domain binds phosphatidylinositol lipids: PI(3,4)P 2 and PI(3,4,5)P 2 and in vivo the PH domain is thought to target these proteins to the plasma membrane in response to extracellular ligands such as PDGF Chang et al., 2012). Although the MRL proteins have been shown to help recruit Ena/VASP proteins to the plasma membrane, their effect on Ena/VASP activity has never been characterized. In addition, results from several studies indicate that the MRL proteins likely have additional, Ena/VASP-independent roles in actin network regulation Lyulcheva et al., 2008;Michael et al., 2010).
To better understand the molecular mechanisms underlying cellular control of actin assembly we characterized the interactions between human VASP, Lpd, and filamentous actin in vitro and in live cells. To our surprise, Lpd binds directly to filamentous actin, both in the absence and presence of VASP, an interaction mediated by a cloud of positively charged basic residues scattered through the C-terminal region of the protein (the Lpd Actin Binding Region (ABR), residues 850-1250). In cells, Lpd 850−1250aa , lacking the RA-PH domain localizes to leading edge membranes and undergoes retrograde flow with the actin cytoskeleton. Surprisingly, the interaction between Lpd 850−1250aa and the actin cytoskeleton does not require interactions with Ena/VASP proteins or SH3 containing proteins, such as Abi1 and endophilin. In vitro, Lpd increases processivity of barbed end-associated VASP tetramers by tethering them to actin filaments. Together these results provide mechanistic insight into how growing actin filaments feed back on the polymerases and nucleation promoting factors that regulate their assembly.

Lpd binds directly to single actin filaments in vitro
Lpd takes its name from the dynamic lamellipodial actin networks to which it localizes in vivo, even in the absence of Ena/VASP proteins or free actin filament barbed ends . This tenacious localization to leading edge actin networks suggested that Lpd might interact directly with actin filaments, so we tested this idea by simultaneously visualizing monomeric GFP-Lpd 850−1250aa and individual actin filaments in vitro by Total Internal Reflection Fluorescence (TIRF) microscopy ( Figure 1A,B). In buffer containing 50 mM KCl, the GFP-labeled, monomeric Lpd construct uniformly decorated actin filaments, with a measured dissociation equilibrium constant (K d ) of 255 ± 2 nM ( Figure 1B,C). Consistent with a weak electrostatic interaction, Lpd binding to filamentous actin grew progressively weaker in buffers containing higher concentrations of KCl. In the presence of 100 mM KCl, interaction between monomeric Lpd and single actin filaments were undetectable by TIRF-M ( Figure 1B). In contrast to these single-filament TIRF assays, we were able to detect interactions between Lpd 850−1250aa and filamentous actin by co-sedimentation in the presence of physiological salt concentrations (Figure 1-figure supplement 1A,B). The stronger actin filament binding observed by co-sedimentation likely results from Lpd bundling actin filaments in solution.
Conserved acidic residues near the amino terminus of actin create an electronegative patch on the surface of actin filaments that interacts with positively charged, basic residues in many actin binding proteins (Fujii et al., 2010). The carboxy-terminal region of human Lpd (residues 850-1250) is highly basic, with an isoelectric point (pI) of 9.97 (Figure 2A,B). The distribution of basic residues in this (C) Calculation of K d for GFP-Lpd 850−1250aa actin filament binding using the average fluorescence intensity of GFP-Lpd bound to phalloidin stabilized actin filaments (20% Cy5 labeled). Error bars represent standard error of the mean. (D) Mutations in all 44 lysine/arginine residues to alanine (called Lpd 44A ) abolish F-actin binding of GFP-Lpd 850−1250aa . Actin filament binding was visualized in the presence 500 nM GFP-Lpd 850−1250aa , wild-type and 44A mutant, in the presence of 50 mM KCl containing buffer as in (B). Scale bar, 10 μm. (E) Purification of GFP-Lpd 850−1250aa (monomer), GFP-LZ-Lpd 850−1250aa (dimer), Figure 1. continued on next page region is evolutionarily conserved among Lpd homologs, but is not conserved in other Ena/VASPbinding proteins, such as Rap1-GTP-interacting adaptor molecule (RIAM), zyxin, or ActA ( Figure 2B). Mutating all 44 basic amino acid residues in the C-terminal ABR of Lpd to alanines (to create Lpd (44A) 850−1250aa ) abolished filament binding ( Figure 1D), but had no effect on the ability of this construct to bind to Ena/VASP proteins (unpublished results). Because the C-terminal region of Lpd appears to be natively unfolded, the loss of actin binding likely reflects simply a loss of electrostatic interactions rather than a change in protein structure. Furthermore, bacterially expressed and purified GFP-Lpd (44A) 850−1250aa showed no signs of being structurally or functionally compromised, as compared wildtype GFP-Lpd 850−1250aa .
Structural studies indicate that full-length (FL) Lpd forms homo-dimers in solution, via both a coiled-coil motif and interactions in the tandem RA-PH domains (Depetris et al., 2009;Chang et al., 2012). We, therefore, tested the effect of Lpd oligomerization on actin filament binding. We created stable dimers of the C-terminal region of Lpd by fusing Lpd 850−1250aa to a dimer-forming, leucinezipper motif (GFP-LZ-Lpd, Figure 1E,F). Unlike the monomeric Lpd constructs, GFP-LZ-Lpd dimers bound to single actin filaments in both low-and high-salt buffers (100 mM KCl; Figure 1G). Similarly, dimeric Lpd bound more strongly to filamentous actin, both 'native' and phalloidin stabilized, as compared to monomeric Lpd in a co-sedimentation assay ( Figure 1H,I and Figure 1-figure supplement 1). Based on the ratio of Lpd:Actin sedimented under saturating conditions, we estimate a stoichiometry of one GFP-Lpd 850−1250aa to at least two actin protomers.
We also purified spontaneously formed trimeric and tetrameric oligomers of his 10 GFP-Lpd 850−1250aa by size-exclusion chromatography and found that their affinity for filamentous actin was even further enhanced ( Figure 1E,F). In our TIRF assay we observed oligomeric his 10 GFP-Lpd 850−1250aa particles bind and diffuse linearly along the sides of actin filaments ( Figure 1J), an activity we previously observed for tetrameric Cy3-VASP constructs (Hansen and Mullins, 2010). Finally, we observed enhanced filament binding when we coupled freshly gel-filtered, monomeric his 10 GFP-Lpd 850−1250aa to small unilamellar vesicles (SUVs) containing DOGS-NiNTA lipids ( Figure 1K). These Lpd-coated SUVs bound tightly to individual actin filaments in vitro, demonstrating that oligomerization and/or clustering of Lpd 850−1250aa promotes stable actin filament binding, even in near physiological salt concentrations.
In addition to slowing network growth, membrane-tethered Lpd also polarized in an actindependent manner on the surface of LCBs. Similar to the polarization of N-WASP on lipid vesicles in Xenopus egg extract (Co et al., 2007), we found that Scar/WAVE and Lpd constructs become enriched in regions of the membrane most closely associated with the growing actin network. In the absence of actin, both his 10 Cherry-SCAR and his 10 GFP-Lpd 850−1250aa were distributed uniformly around the LCBs ( Figure 3D). Within 10 min of initiating actin network assembly on LCBs, however, both membrane-tethered Scar/WAVE and Lpd constructs (his 10 GFP-Lpd 850−1250aa ) became concentrated in the region adjacent to the actin network, a region we refer to as the 'barbed end attachment zone' ( Figure 3E, Figure 3-figure supplement 1B-D). In contrast, membrane-tethered his 10 GFP was displaced away from the barbed end attachment zone and became concentrated on the opposite side of the microsphere ( Figure 3F).

Lpd interacts with the actin cytoskeleton in vivo
Having found that Lpd binds actin filaments in vitro, we next worked to determine whether actin binding by Lpd plays a significant role in its cellular localization and function. To address this question we visualized the localization of fluorescently tagged, FL Lpd (GFP-Lpd 1−1250aa ) in Xenopus Tissue Culture (XTC) cells, spread on poly-L-lysine (PLL)-coated coverslips ( Figure 4A). We used XTC cells because they spread well and produce very thin peripheral actin networks, well suited for fluorescence microscopy. Ectopic gene expression from a truncated cytomegalovirus (CMV) promoter produces very low levels of protein expression in XTC cells, ideal for single-molecule fluorescence microscopy (Watanabe and Mitchison, 2002). TIRF microscopy imaging of single GFP-Lpd 1−1250aa molecules revealed that FL Lpd localizes predominantly to the leading edge of ruffling membranes and cycles on and off the plasma membrane on a subsecond time scale ( Figure 4B, Figure 4-figure supplement 1, Video 1). Leading edge membrane localization of GFP-Lpd 1−1250aa required dynamic actin assembly and disassembly, because acute treatment of cells with a chemical inhibitor cocktail that freezes actin dynamics (JLY drug cocktail: Jasplakinolide, Latrunculin B, and Y27632 Rock kinase inhibitor [Peng et al., 2011]) caused rapid loss of membrane-localized Lpd ( Figure 4C).
To determine which Lpd binding partners-actin, Ena/VASP, Abi1/endophilin, inositol phospholipids, or Ras/Rho-family G-proteins-are required for leading edge localization, we compared the localization of various Lpd truncation mutants in XTC cells. We discovered that not only does GFP-Lpd 850−1250aa , which lacks the tandem RA-PH domains, localize to leading edge membranes in XTC cells ( Figure 4C), this construct also undergoes retrograde flow with the lamellipodial actin network. Analysis of speckle velocities in XTC cells co-expressing GFP-Lpd 850−1250aa and mCherry-Actin revealed that the two proteins move with the same mean velocity: 71.9 ± 17.5 nm/s for Lpd 850−1250aa vs 73.5 ± 14 nm/s for actin ( Figure 4F, Video 2). In contrast to mCherry-Actin, GFP-Lpd 850−1250aa speckles were observed less frequently, appeared more diffuse, and had shorter lifetimes (Figure 4-figure supplement 2). We suspect that the larger GFP-Lpd 850−1250aa fluorescent particles are associated with fast endophilin-mediated endocytosis (FEME, Vehlow et al., 2013;Boucrot et al., 2015). Consistent with constitutively dimeric Lpd having a higher affinity for actin, GFP-LZ-Lpd 850−1250aa displayed slightly Arp2/3, 100 nM Mm capping protein, and 3 μM hCofilin. (E) his 10 Cherry-SCAR APWCA and his 10 GFP-Lpd 850−1250aa concentrate on the barbed end dense side of the actin comet tail. (F) his 10 -Cherry-SCAR APWCA concentrates on the barbed end dense side of the actin comet tail, while his 10 -GFP is excluded from the barbed end attachment zone. Line scans across LCBs are shown to the right. Scale bar, 5 μm. DOI: 10.7554/eLife.06585.006 The following figure supplement is available for figure 3:   We also transiently expressed GFP-Lpd 850−1250aa in polarized B16F1 mouse melanoma cells migrating on laminin coated glass, and observed the same leading edge membrane localization as in XTC cells ( Figure 4G, Figure 4-figure supplement 1, Video 3). Importantly, this localization does not require acute receptor stimulation following serum starvation, implying that membrane recruitment of GFP-Lpd 850−1250aa represents the default Lpd localization in these migrating cells.
We next asked whether GFP-Lpd 850−1250aa requires Ena/VASP binding for leading edge membrane localization in XTC cells. The C-terminal region of Lpd contains six FPPPP peptide sequences, each of which can potentially bind one VASP EVH1 domain. By analytical ultracentrifugation, we found that GFP-Lpd 850−1250aa simultaneously binds up to four VASP EVH1 domains in solution (Figure 5-figure supplement 1). We mutated all six FPPPP motifs in Lpd 850−1250aa to AAPPP (Figure 5-figure supplement 2) and expressed this mutant protein as either a his 10 -GFP fusion in Escherichia coli (his 10 -GFP-Lpd(AAPPPx6) 850−1250aa ) or a GFP fusion in XTC cells (GFP-Lpd (AAPPPx6) 850−1250aa ). Although the his 10 -GFP-Lpd(AAPPPx6) 850−1250aa mutant failed to recruit Ena/VASP proteins to LCBs in vitro ( Figure 5A), the localization of the GFP-Lpd (AAPPPx6) 850−1250aa mutant in XTC cells was nearly indistinguishable from GFP-Lpd 850−1250aa ( Figure 5B). The only noticeable difference was a slight decrease in the amount of protein associated with nascent focal adhesions ( Figure 5B). Video 2. Retrograde flow of GFP-Lpd 850−1250aa and mCherry-Actin in XTC cells spread on poly-L-lysine (PLL) coated coverslips. Cell was imaged with a temporal resolution of 5 s using TIRF-M at 23˚C. GFP-Lpd 850−1250aa and mCherry-Actin were ectopically expressed from either a CMV or DeltaCMV promoter, respectively. Video plays at 10 frames per second. Scale bar, 5 μm. DOI: 10.7554/eLife.06585.012 FL human Lpd 1−1250aa contains 10 potential SH3 domain binding sites , of which four fall within our ABR construct, Lpd 850−1250aa (Figure 2A). To abolish the known interactions with Abi1 and endophilin, we mutated essential proline residues (e.g., PxxPxR → GxxGxR) in all four SH3 domain binding sites ( Figure 5-figure supplement 2). This mutant, termed Lpd(SH3*) 850−1250aa , could still localize to the leading edge in XTC cells ( Figure 5B), but to a lesser extent than wild-type Lpd or the (AAPPP) x6 Lpd mutant which does not bind Ena/VASP EVH1 domains. Combining the (AAPPP) x6 and SH3* mutations in the same protein further reduced the percentage of cells with leading edge membrane localization of these GFP-Lpd constructs ( Figure 5B). In each case, mutations in the Ena/VASP or Abi1/ endophilin binding sites resulted in a loss of leading edge localization that coincided with an increased cytoplasmic localization of the GFP-Lpd 850−1250aa construct.
To test whether the cloud of basic residues in the C-terminal region of Lpd mediates interactions with the actin cytoskeleton in vivo, we created a mutant lacking lysine and arginine residues that flank the multiple Ena/VASP and Abi1/endophilin binding sites (Figure 5-figure supplement 2). In vitro, we found that GFP-Lpd (44A) 850−1250aa lacking all lysine and arginine residues failed to bind to filamentous actin in vitro ( Figure 1D). Because several of these point mutations flank or fall within the SH3 binding sites, we also generated a GFP-Lpd(35A) 850−1250aa actin binding mutant, which lacks only the basic residues that sit between Ena/VASP and Abi1/ endophilin binding sites (Figure 5-figure supplement 2). Similar to GFP-Lpd(44A) 850−1250aa , the GFP-Lpd(35A) 850−1250aa construct also failed to concentrate at actin-based membrane protrusion or nascent focal adhesions in XTC cells ( Figure 5B). We attempted to increase the avidity of GFP-Lpd(35A) 850−1250aa for the leading edge membrane by expressing the protein as a leucine zipper-mediated constitutive dimer (GFP-LZ-Lpd (35A) 850−1250aa ), but this mutant remained cytoplasmic ( Figure 5B).
To determine the role of membrane association in the leading-edge localization of Lpd we attempted to rescue leading edge localization of various GFP-Lpd 850−1250aa mutants by constitutively tethering each protein to the plasma membrane with a Lyn palmitolyation sequence derived from Src kinase ( Figure 5C) (Inoue et al., 2005). Similar to what has previously been observed for membrane tethered ActA-(FPPPP) x4 -CAAX  and soluble GFP-Lpd 850−1250aa , described here, Lyn-GFP-Lpd 850−1250aa localized to the leading edge of actin-based membrane protrusions ( Figure 5D, Video 4). In contrast, leading edge localization of the combined (AAPPP) x6 and SH3* mutant construct was observed in fewer cells than the Lyn-GFP-Lpd 850−1250aa . Although Lyn-GFP-Lpd(35A) 850−1250aa was targeted to the plasma membrane by the palmitoyl lipid anchor, this mutant still failed to localize to the peripheral regions where Ena/VASP, Abi1, and endophilin are located ( Figure 5D). Localization of Lyn-GFP-Lpd(35A) 850−1250aa was identical to that of a membrane-anchored Lyn-GFP construct lacking any Lpd sequences. Both of these constructs were distributed uniformly across the plasma membrane ( Figure 5D).
Given this loss of positive charges, it is possible that failure of Lyn-GFP-Lpd (35A) to localize to the leading edge membrane reflects an inability to interact with a polarized population of anionic lipids.
Video 3. Leading edge membrane localization of GFP-Lpd 850−1250aa in polarized B16F1 mouse melanoma cell migrating on laminin coated glass. B16F1 cell was imaged with a temporal resolution of 10 s using widefield epi fluorescence at 37˚C. GFP-Lpd 850−1250aa was ectopically expressed from a CMV promoter. Video plays at 10 frames per second. Scale bar, 20 μm. DOI: 10.7554/eLife.06585.013  To look for such a polarized lipid distribution we expressed a phosphatidylinositol lipid sensor, based on the pleckstrin-homology domain of phospholipase Cδ (GFP-PLCδ PH). We found that GFP-PLCδ PH distributes uniformly across the plasma membrane of XTC cells ( Figure 5D). This result supports our interpretation that failure of Lyn-GFP-Lpd (35A) to localize to leading edge membranes in XTC cells reflects the fact that Lpd coupling to the actin cytoskeleton is required for establishing interactions with Ena/VASP or Abi1/ endophilin at actin based membrane protrusions.
To test whether leading edge membrane localization of GFP-Lpd 850−1250aa requires interaction with free actin filament barbed ends or with the sides of actin filaments, we used pharmacological perturbations to examine how GFP-Lpd 850−1250aa leading edge localization responds to changes in actin cytoskeletal dynamics. First, we simply froze actin assembly and disassembly by adding a mixture of Jasplakinolide and Latrunculin B to cells, without inhibiting myosin activity. Under these conditions we observed rapid retrograde movement of GFP-Lpd 850−1250aa away from the leading edge, identical to the movement of mCherry-Actin ( Figure 6A, Video 5). The dynamics of GFP-Lpd 850−1250aa translocation suggests that, in the absence of net actin assembly, myosin can pull GFP-Lpd 850−1250aa off the leading edge via a tight linkage to the actin cytoskeleton. In these experiments, a fraction of membrane-localized GFP-Lpd 850−1250aa remained at the leading edge, presumably due to the maintenance of interaction with Ena/VASP and/or Abi1/endophilin.
Next we treated cells expressing GFP-Lpd 850−1250aa with Cytochalasin D, a small molecule that caps barbed ends of actin filaments. Consistent with previous observations Neel et al., 2009), addition of Cytochalasin D to XTC cells strongly inhibited the localization of GFP-VASP to leading edge membranes (unpublished observations). When we treated these cells with a 100 nM Cytochalasin D, we observed a large fraction of leading edge-localized GFP-Lpd 850−1250aa immediately dissociate from the leading edge membrane and flow back with the actin cytoskeleton ( Figure 6B, Video 6). Because GFP-LZ-Lpd 850−1250aa binds more strongly to filamentous actin, dissociation from the leading edge membrane was more dramatic following addition of Cytochalasin D ( Figure 6C, Video 7). The same membrane dissociation dynamics were observed when cells expressing GFP-Lpd 850−1250aa mutants, (AAPPP) x6 and SH3*, were treated with Cytochalasin D ( Figure 6D,E). Together, these results suggest that the leading edge localization of GFP-Lpd 850−1250aa is regulated, in part, by a direct interaction with free actin filament barbed ends, independent of interactions with Ena/VASP or Abi1/endophilin proteins.

Lpd can tether VASP to actin filaments
The six VASP-binding 'FPPPP' motifs in the C-terminal region of Lpd do not appear to overlap the cloud of basic residues that mediate actin filament binding (Figure 2A). We, therefore, wondered whether Lpd could bind Ena/VASP proteins and filamentous actin simultaneously. To test this idea, we first used TIRF microscopy to demonstrate that several Cy3-labeled VASP constructs-an isolated EVH1 domain and two full length, tetrameric, actin-binding mutants-failed to bind actin filaments ( Figure 7A). Addition of GFP-Lpd 850−1250aa , however, caused all of these VASP constructs to associate with the sides of actin filaments ( Figure 7B). Interaction of the tetrameric Cy3-VASP actin-binding mutant with the monomeric Lpd construct produced higher-order VASP clusters visible on the sides of the filaments (Figure 7B, arrowheads). We previously observed this type of protein clustering in vitro when we combined tetrameric VASP and ActA 255−392aa (unpublished observations).
Supporting the notion that Lpd-VASP complexes can synergistically bind actin filaments, we find that the presence of both VASP and GFP-LZ-Lpd results in a large amount of actin filament bundling in the presence of 100 mM KCl (Figure 7-figure supplement 1). In contrast, individual actin filaments

Lpd enhances VASP barbed end processivity
Because human VASP is a weakly processive actin polymerase and Lpd can interact simultaneously with both VASP tetramers and actin filaments, we hypothesized that Lpd might modulate the processivity of VASP by both clustering tetramers and tethering them to actin filaments. We tested this hypothesis by visualizing actin filament elongation in the presence of either monomeric (Lpd 850−1250aa ) or dimeric (LZ-Lpd 850−1250aa ) GFP-Lpd constructs. The weak interaction between monomeric GFP-Lpd 850−1250aa and filamentous actin turns out to be further antagonized by the presence of monomeric actin or profilin-actin ( Figure 8A). Also, we observed a reduction in the rate of barbed end actin filament elongation in the presence 2 μM profilin-actin and soluble GFP-Lpd 850−1250aa or GFP-LZ-Lpd 850−1250aa ( Figure 8B), consistent with Lpd binding to (and partially sequestering) actin monomers. In contrast to monomeric Lpd, dimeric GFP-LZ-Lpd 850−1250aa could transiently interact with both the sides and barbed ends of single growing actin filaments ( Figure 8D, Video 8). Although GFP-Lpd and GFP-LZ-Lpd associate more transiently with single actin filaments in the presence of monomeric actin, Lpd binds well enough to promote filament bundling when present at a near stoichiometric concentration relative to monomeric actin (Figure 8-figure supplement 1).
We tried to measure the barbed end dwell time of GFP-LZ-Lpd 850−1250aa , but the dissociation kinetics and diffusivity of Lpd were too rapid. Using lower concentrations of GFP-LZ-Lpd 850−1250aa to visualize single barbed end association events did not improve our resolution. We speculate that the comet-like barbed end localization of GFP-LZ-Lpd 850−1250aa ( Figure 8D) may reflect dimers simultaneously binding to the barbed end and near the barbed end, creating a fluorescent speckle.
Because GFP-Lpd and GFP-LZ-Lpd reduce the rate of actin filament elongation and transiently associate with barbed ends, we tested whether VASP could alleviate the inhibitory effect of Lpd. We measured barbed end elongation rates in the presence of 2 μM actin and 2 μM profilin-actin, as well as buffers of various ionic strengths ( Figure 8B,C). Because Lpd and VASP synergistically and potently bundle actin filaments in 50 mM KCl, we tried increasing the ionic strength (75-100 mM KCl) to minimize side binding, while favoring barbed end association of the Lpd-VASP complex. However, for all conditions tested, we found that the rate of barbed end elongation in the presence of both Lpd and VASP was intermediate between the rates measured in the presence of either Lpd or VASP alone ( Figure 8B,C). Interpreting these results remains challenging because multiple competing activities contribute to single filament elongation rates.
One challenge we faced in visualizing the colocalization of Lpd and VASP on single actin filament barbed ends was that the high solution protein concentrations required to observe complex formation. Unfortunately, these high concentrations were not compatible with single molecule imaging and produced large protein complexes that were unable to associate with growing barbed ends. To circumvent this problem, we used lower concentrations of his 10 GFP-Lpd (850-1250aa) oligomers purified by size exclusion chromatography ( Figure 1E) and characterized their effect on VASP barbed end processivity in the single actin filament TIRF assay. Under these conditions we observed complexes containing both VASP and Lpd processively track growing actin filament barbed ends at a velocity of 36 ± 9.4 subunits/s ( Figure 8I, Video 9). The rate at which Lpd-VASP complexes shuttled actin monomers onto free barbed ends was faster than elongation of single actin filaments in the presence of 50 nM tetrameric VASP (25 ± 1.6 subunits/s, Figure 8I). The dwell time for VASP-Lpd tip complexes (τ 1 = 33 ± 2 s; Figure 8J) was also significantly longer as compared to single VASP tetramers (τ 1 = 0.49 ± 0.03 s; Figure 8F).

Lpd binds directly to actin filaments
We discovered a direct interaction between Lpd and filamentous actin that is mediated by a highly basic, unstructured region in the C-terminus of Lpd (residues 850-1250). The cloud of positively-charged amino acid residues is conserved between Lpd homologs of different organisms. These residues, however, are absent from other Ena/VASP binding proteins, such as RIAM, ActA, and Zyxin ( Figure 2B), suggesting that this basic region is important for cellular functions specific to Lpd. The filament-BD does not overlap motifs that bind the EVH1 domains of VASP (Figure 2A), enabling Lpd to simultaneously interact with both Ena/VASP proteins and filamentous actin. Similar to VASP (Hansen and Mullins, 2010), monomeric Lpd molecules bind actin filaments only weakly in physiological ionic strength (∼150 mM KCl), but bind much more strongly when oligomerized or densely clustered on membranes. Because of this effect, local enrichment of Lpd at the plasma membrane can exert a non-linear, cooperative effect on Ena/VASP activity (Figure 9).
Several mechanisms have been proposed to modulate the activity of Ena/VASP proteins, including: post-translational modifications of the EVH2 domain (Barzik et al., 2005); alternative splicing of Ena/ VASP transcripts (Philippar et al., 2008); and specificity for distinct profilin-actin isoforms (Dugina et al., 2009;Mouneimne et al., 2012). In this study, we establish that Ena/VASP proteins are directly controlled by their membrane-targeting factors. Specifically we show that Lpd-VASP complexes are more persistently localized to polymerizing ends, because the Lpd-actin interaction tethers VASP to the barbed end (Figure 9). Filament tethering in combination with VASP clustering (Breitsprecher et al., 2008) greatly increases the processivity of Ena/VASP proteins. In addition, Lpd-VASP complexes also enhance barbed-end polymerization more effectively compared to individual VASP tetramers, an ability likely based on Lpd's interaction with actin monomers. By increasing the number of monomer binding sites near the barbed end, Lpd may directly contribute to the polymerase activity of VASP tetramers.
In cells both FL Lpd and its C-terminal actin BD (Lpd 850−1250aa ) localize to actin-based membrane protrusions independently of other binding proteins such as Ena/VASP or Abi1/endophilin. In contrast to FL Lpd, however, we find that acute treatment of cells with cytochalasin D rapidly displaces GFP-Lpd 850−1250aa from the leading edge and causes it to translocate toward the cell body with the disassembling lamellipodial actin network. We observed a similar phenotype for Lpd 850−1250aa mutants that cannot interact with Ena/VASP proteins or SH3 domains. Taken together these results demonstrate that the C-terminal region of Lpd harbors autonomous filament binding sites that strongly influence its localization to actin rich protrusions. The ability of Lpd 850−1250aa to directly interact with polymeric actin extends the canonical model of Lpd membrane localization, which suggests that Lpd targets the plasma membrane by binding to PI(3,4)P 2 and/or small GTPases . Future experiments will determine the hierarchical mechanisms that control membrane localization and maintenance of Lpd in response to these diverse molecular inputs.

Role of profilin in modulating Lpd actin binding
In vitro, the C-terminal actin BD of Lpd reduces barbed end elongation by partially sequestering actin monomers and/or transiently capping free barbed ends. This effect is more pronounced in the presence of profilin-actin, as compared to actin alone, indicating that profilin can shunt Lpd towards the filament barbed end. Interestingly, Lpd contains a poly-proline domain situated upstream of the ABR characterized in this study. We speculate that this poly-proline domain, similar to its function in other actin regulatory proteins such as formins and nucleation promoting factors (Paul et al., 2008;Campellone and Welch, 2010), could bind profilin-actin and promote its interaction with the actinbinding region, influencing the partitioning of Lpd between monomeric and filamentous actin. Previous work by Bae et al. (2010) indicates that the interaction between profilin and phosphatidylinositol lipids (e.g., PI(3,4)P 2 ) plays a key role in regulating Lpd localization to actinbased membrane protrusions. It will, therefore, be important to characterize the interplay between profilin binding to the Lpd poly-proline domain and phosphatidylinositol lipids in the future. We speculate that enrichment of PI(3,4)P 2 at the plasma membrane could liberate monomeric actin from profilin-actin complexes, thereby promoting the interaction between Lpd and actin monomers.

The role of Lpd in actin network assembly
The function of Lpd at the leading edge is not simply to recruit Ena/VASP proteins to the plasma membrane: Knockdown of Lpd inhibits lamellipodium formation more potently compared to loss of Ena/VASP proteins Krause et al., 2004) and results in global reduction of lamellipodial actin density (Lyulcheva et al., 2008;Michael et al., 2010;Law et al., 2013). Recently both Lpd and its binding partners Ena/VASP were found to interact with the Scar/WAVE regulatory complex Chen et al., 2014), an important activator of the Arp2/3 complex at the leading edge. These interactions are mediated by the SH3 domain of Abi1, a WAVE complex component, which binds multiple PxxPxR motifs in the C-terminus of Lpd and the EVH1 domains of Ena/VASP proteins. Mutations interfering with WAVE-Lpd binding impaired cell migration in wound closure assays , demonstrating the functional relevance of this interaction. Direct binding of Lpd to filamentous actin might promote Ena/VASP and Scar/WAVE activity within the lamellipodial actin networks they construct. The position of the WAVE-binding motif in Lpd suggests that Lpd could synchronously tether the WAVE regulatory complex and Ena/VASP to filamentous actin. Furthermore, both Lpd Chang et al., 2012) and the WAVE regulatory complex interact with anionic lipids at the plasma membrane (Lebensohn and Kirschner, 2009;Koronakis et al., 2011). We propose that these multi-layered interactions at the leading edge guide the assembly of the actin nucleation and elongation machinery into supramolecular complexes to coordinate their activities for the efficient construction of force-generating actin networks.
Besides its function in regulating lamellipodial actin network assembly, Lpd has also been implicated in regulating FEME Boucrot et al., 2015). Similar to the interactions with the WAVE complex, the C-terminus of Lpd binds to SH3 domains in endophilin using multiple PxxPxR motifs. Consistent with Lpd playing a role FEME, we observed vesicular localization of GFP-Lpd 850−1250aa that translocate toward the cell body. Because Lpd interacts directly with filamentous actin and multiple actin regulatory proteins, Lpd could function to orchestrate the actin nucleation machinery at the sites of FEME.

Molecular biology
Lpd clones were derived from a plasmid encoding human Lpd 1−1250aa (pBSII-SK(+) vector) provided by Matthias Krause (King's College London). his 10 -TEV-EGFP (A207K) -Lpd 850−1250aa was cloned into a modified pETM-11 vector (EMBL Protein Expression and Purification Facility; Heidelberg, Germany). Expression vectors used for transient transfection into XTC cell, genes were cloned into the pCMV-EGFP/mCherry (Takara Clontech Laboratories, Mountain View CA, USA) or pDeltaCMV-EGFP vectors originally created by Naoki Watanabe (Tohoku University, Japan) and provided by Orion Weiner (University of California at San Francisco, CVRI). All of our constructs contain a monomeric EGFP variant, A207K point mutation, which significantly reduces the tendency of EGFP to dimerize. Inspired by experiments described by Inoue et al. (2005), we fused a Lyn 11 palmityolation sequence (GCIKSKGKDSA) derived from Src family kinase to our pDeltaCMV-EGFP(A207K) construct to generate our Lyn-GFP and Lyn-GFP-Lpd fusions. Mutations in the Lpd Ena/VASP BDs, (AAPPP) x6 , were introduced using Pfu Turbo DNA polymerase and standard site-directed mutagenesis techniques. We made the constitutive GFP-Lpd dimer by inserting a leucine zipper derived from Saccharomyces cerevisiae Gcn4p; (Tomishige et al., 2002) in between the GFP and Lpd protein sequences [EGFP-GGGYKQLEDKVEELASKNYHLENEVARLKKLVEFGGG-Lpd]. Lpd 850−1250aa with all lysine and arginine residues mutated to alanine, called Lpd (44A), was synthesize by Invitrogen (GeneArt) and cloned into the pDeltaCMV-Lyn-EGFP(A207K) or his 10 -TEV-EGFP(A207K) vectors for transient cell expression and bacterial protein expression, respectively. We used the Quikchange Lightning Multi site-directed mutagenesis kit (Agilent, Cat# 210515) to correct lysine/arginine to alanine mutations that flanked or lied within the Abi1/endophilin SH3 domain binding sites resulting in a construct we termed Lpd (35A). This kit was separately used to mutate proline residues within the four predicted SH3 domain-binding sites located in Lpd 850−1250aa . The pIs for Lpd and homologs were calculated with EXPASY ProtParam (Wilkins et al., 1999) and sequences were aligned using ClustalW. Refer to Figure 5-figure supplement 2 for exact protein sequences of each Lpd 850−1250aa mutant used in this study. For a complete list of plasmid DNA used in this study, refer to Supplementary file 1.
Removal the his 6 -Ztag or his 10 , was achieved by TEV protease digestion of the HiTrap eluate during an overnight dialysis in 4 liters of buffer containing 50 mM Na 2 PO 4 [pH 8.0], 400 mM NaCl, 0.4 mM BME. After 20-24 hr, the dialysate containing TEV cleaved Lpd was recirculated over a HiTrap chelating column to remove uncleaved protein, his 6 -TEV protease, and other proteins that nonspecifically bound during the first purification step. EGFP-Lpd and EGFP-LZ-Lpd were then in the presence of 200 nM GFP-LZ-Lpd 850−1250aa (τ 1 = 0.58 ± 0.05 s (73%, fast), τ 2 = 2.3 ± 0.4 s (27%, slow), n = 632 molecules). Note that the dwell times for Cy3-VASP in (F) are shorter than previously reported (Hansen and Mullins, 2010). This due to Cy5-Actin being a less favorable substrate for barbed end incorporation compared to Alexa488-Actin. (H) Clustered his 10 -GFP-Lpd 850−1250aa increases the processivity of Cy3-VASP. Image montage showing colocalization of the Cy3-VASP (5 nM) and his 10 -GFP-Lpd 850−1250aa (50 nM) on actin filament barbed end elongating in the presence of 2 μM Actin (20% Cy5) and TIRF buffer contains 75 mM KCl. Note the intensity of the actin filament decreases when the VASP-Lpd complex is associated with the growing actin filament barbed end, indicating that unlabeled vs Cy5-labeled actin is more favorably incorporated. Scale bar, 5 μm. (I) Lpd-VASP barbed associated complexes incorporate actin monomers at a faster velocity, as compared to actin filament elongating in the presence of 50 nM tetrameric VASP. Error bars represent the standard deviation of the mean (p-value = 7 × 10 −12 ; two-tailed t-test for data sets with unequal variance). (J) Calculation of the barbed end dwell times for Cy3-VASP and his 10 -GFP-Lpd 850−1250aa complexes. Plot of 1-CDF was best fit to a single exponential curve, yielding τ 1 = 33 ± 2 s (n = 87 complexes). DOI: 10.7554/eLife.06585.025 The following figure supplement is available for figure 8: To isolate uncleaved his 10 -TEV-EGFP(A207K)-Lpd 850−1250aa , we loaded the HiTrap eluted protein onto a 323 ml Superdex200 gel filtration column equilibrated with 10 mM HEPES [pH 7.0], 200 mM NaCl, 1 mM TCEP. This allowed us to separate soluble aggregates from the monomeric protein. Peak fractions were then combined with 10 mM HEPES [pH 7.5], 1 mM TCEP (1:1 vol) to effectively reduce the salt concentration to 100 mM NaCl. This sample was then loaded on a MonoS column equilibrated with 10 mM HEPES [pH 7.0], 100 mM NaCl, 1 mM TCEP and gradient eluted as described above. Peak fractions of his 10 -EGFP-Lpd eluted from the MonoS column containing a small fraction of spontaneously formed oligomers. Freezing in glycerol, sucrose, arginine-glutamate, or without cryo-protection also caused his 10 -EGFP-Lpd protein to oligomerize. Consequently, his 10 -EGFP-Lpd was frozen after MonoS elution as 300-500 μl aliquots (40-50 μM). When needed for actin bead motility assays and TIRF microscopy experiments, his 10 -EGFP-Lpd was thawed and gel filtered using a 24 ml Superdex200 column equilibrated with 10 mM HEPES [pH 7.5], 200 mM NaCl, and 1 mM TCEP. Freshly gel filtered his 10 -EGFP-Lpd showed no signs to proteolysis or aggregation after at least one week on ice.  , Lpd is recruited to actin based membrane protrusions through interactions with phosphatidylinositol lipids (i.e., PI(3,4)P 2 ) and possibly small GTPases (i.e., Ras or Rho family). Similar to the Grb protein family, Lpd is predicted to form homo-dimers mediated by interactions between the coiled-coil and tandem RA-PH domain. We find that the C-terminus of Lpd (residues 850-1250) is sufficient for recruiting Lpd to leading edge membrane where it directly interacts with free barbed ends and/or the sides of the actin filaments. Importantly, this interaction between Lamelliopodin and filamentous actin can occur independently to those mediated by Ena/VASP proteins or SH3 domains (i.e., Abi1/endophilin). However, Ena/VASP proteins recruited to actin based membrane protrusion can simultaneously associate with free actin filament barbed ends and Lpd. By this mechanism, we speculate that the lifetime of membrane targeted and barbed end associated Ena/VASP proteins are extended at the plasma membrane. DOI: 10.7554/eLife.06585.027 Cytoplasmic actin was purified from Acanthamoeba castellani as described (Gordon et al., 1976;Hansen et al., 2013). Actin was stored at 4˚C in G-buffer containing 2 mM Tris [pH 8.0], 0.5 mM TCEP, 0.1 mM CaCl 2 , 0.2 mM ATP, 0.01% azide, and used within 6 months. The quality, or degree of proteolysis, of monomeric actin was routinely monitored by SDS-PAGE. Monomeric actin was labeled on Cys-374 by adding 3-7 molar excess Cy3-maleimide (GE Healthcare, Cat# PA23031) or Cy5maleimide (GE Healthcare, Cat# PA25031) on ice in G-buffer for 10-15 min. Reactions were quenched with 10 mM DTT and then centrifuged at 346716×g (TLA 100.4 rotor, Beckman Coulter) to remove insoluble dye and aggregated protein. Labeled actin was then polymerized at room temperature by the addition of KMEI polymerization buffer (final concentration of 10 mM imidazole [pH 7.0], 50 mM KCl, 1 mM MgCl 2 , 1 mM EGTA, and 1 mM ATP). Labeled filamentous actin was then centrifuged in a TLA 100.4 rotor for 30 min at 195028×g. The pellet was gently washed with G-buffer and then resuspended in G-buffer (2 mM Tris [pH 8.0], 0.5 mM TCEP, 0.1 mM CaCl 2 , 0.2 mM ATP, 0.01% azide) to initiate actin filament depolymerization. Following depolymerization in G-buffer for 3-5 days, actin was centrifuged at 346716×g (TLA 100.4 rotor) for 20 min and then gel filtered (Superdex 75) in Gbuffer. We typical observed a 50-60% labeling efficiency.
The Arp2/3 complex (native, A. castellani), capping protein (mouse, recombinant), and profilin (human, recombinant) were purified and handled as previously described by Akin and Mullins (2008). Purification of his 10 -Cherry-SCAR APWCA (171-559aa, human WAVE1) was achieved by expressing codon optimized Z-tag-TEV-his 10 -Cherry-SCAR APWCA in Rosetta (DE3) bacteria for 20 hrs at 18˚C in Terrific Broth media. Cell lysate was recirculated over a 5 ml HiTrap chelating column charged with CoCl 2 . Eluted protein was cleaved with TEV protease overnight at 4˚C and then exchanged into low ionic strength buffer using a HiPrep Desalting 26/10 G25 Sephadex column. FL his 10 -Cherry-SCAR APWCA was then separated from partially translated fragments using anion exchange chromatography (i.e., MonoQ) and then separated from protein aggregates by size exclusion chromatography (Superdex75 column).

Analytical ultracentrifugation
For sedimentation equilibrium experiments, GFP-Lpd 850−1250aa and VASP 1−114aa (monomeric EVH1 domain) were diluted into 20 mM HEPES [pH 7], 100 mM KCl, 1 mM TCEP and then centrifuged at 346716×g (TLA 100.4 rotor, Beckman Coulter) for 20 min to remove proteins that were potentially aggregated. GFP-Lpd 850−1250aa (9-10 μM) was then combined with 10, 25, 50, 75, or 100 μM VASP 1−114aa (13.1 kDa) and loaded into 6-well Teflon chambers with quartz windows and placed in a 4 channel Ti-60 rotor. Proteins were centrifuged at 7000, 10,000, and 14,000 rpm in a Beckman XL-I analytical ultracentrifuge at 20˚C for 14-18 hr per speed or until equilibrium was reached. Continuous scans of 488 nm absorbance were acquired every 2 hr in replicates of 10, to monitor the sedimentation of GFP-Lpd 850−1250aa in the absence and presence of VASP 1−114aa . An extinction coefficient of 55,000 M −1 cm −1 was used to calculate the GFP-Lpd 850−1250aa protein concentration from the absorbance measured at 488 nm for each radial position. Global fitting of three equilibrium traces from all speeds (i.e., 7, 10, 14K rpm) for each condition was performed using open source NIH Sedphit and Sedphat software (Peter Schuck, NIH). Equilibrium traces were globally fit using a monomer-dimer self-association model. Using Sednterp, we calculated a partial specific volume of 0.7354 ml/g and a buffer density of 1.00499 g/ml. Our method for fitting the equilibrium traces involved fixing the meniscus position, floating the bottom position, and varying local concentrations in the experimental parameters. Using Sedphat, the apparent molecular weight of GFP-Lpd 850−1250aa was calculated by fixing the dimer concentration at zero and floating the molecular weight.

Single actin filament TIRF assay
Counterglass slides placed in coplin jars were bath sonicated in 3 M NaOH. Glass was rinsed with water, bath sonicated in 100% ethanol, and then dried. To minimize protein depletion in the TIRF experiments, the counterglass was coated with PLL-PEG (PLL-PEG), which was dried onto the glass surface before rinsing with water. Synthesis of PLL-PEG was performed as previously described by Huang et al. (2001). PEG/biotin-PEG functionalized coverslips were then attached to PLL-PEG coated counterglass with double sided tape (Tesa). To reduce nonspecific binding to the PEG functionalized surface the imaging chamber was washed with a 1× PBS solution pH 7.2 containing 1% Pluronic F-127 and 100 μg/ml beta casein. Glass was then washed with buffer containing 20 mM HEPES pH 7, 1 mM DTT, 200 mM KCl, 100 μg/ml beta-casein (wash buffer). Glass was subsequently incubated for 2 min with 50 nM streptavidin, followed by 50 nM biotin-PEG 11 heavy meromyosin diluted in wash buffer. To minimize nonspecific binding of GFP-Lpd and actin, it is essential that the stock solution of kappa casein is made fresh in water and centrifuged for 20 min at 346716×g (TLA 100.4 rotor or equivalent) to remove aggregated protein. If this is not possible, kappa casein can be left out of the 1% Pluronic F-127 or replaced with beta casein. Ultimately, the assay quality will depend on your ability to block defects on the biotin-PEG functionalized glass surfaces. Because Lpd 850−1250aa and human VASP are highly basic proteins, poor coverslip functionalization will result in a tremendous amount of non-specific protein absorption to the glass substrates used for TIRF microscopy.
To calculate the dissociation rate constant (K d ) for GFP-Lpd 850−1250aa binding to single actin filaments ( Figure 1C) we used TIRF microscopy to image the density of GFP-Lpd 850−1250aa bound to phalloidin stabilized actin filaments (20% Cy5 labeled). For these experiments, increasing concentrations of monomeric GFP-Lpd 850−1250aa (0-1 μM) diluted in TIRF buffer containing 50 mM KCl were sequentially flowed into an imaging chamber containing immobilized actin filaments. After a 5 min incubation, GFP-Lpd 850−1250aa and Cy5-Actin filaments were imaged across ≥10 fields of view and ≥2 chambers. The average GFP-Lpd fluorescence intensity was calculated across ≥100 μm of filamentous actin. an eppendorf tube. Diluted glass beads were vortexed and then bath sonicated for 5 min. Monodisperse glass beads were then combined with 25 μl of 4 mM SUVs (4 mM = total lipid concentration), vortexed briefly, and then rotated at room temperature for 30 min. After assembling the lipid bilayer, 750 μl of milliQ water was added to each tube and beads were micro-centrifuged for 2 min at 200×g. The supernatant was aspirated off the beads and the 750 μl water wash, followed by micro-centrifuged was repeated four times. After the final spin, the supernatant was aspirated leaving ∼50 μl of water/beads. Beads were then resuspended by vortexing and 150 μl of buffer containing, 20 mM HEPES [pH 7], 200 mM KCl was added. The final bead slurry was ∼1% (wt/vol) and contained a final KCl concentration of 150 mM.
To charge lipid coated glass beads with protein, we combined 5 μl of 1% bead slurry with 45 μl of 111 nM his 10 tagged protein (i.e., his 10 Cherry-SCAR). Proteins were diluted into buffer containing 20 mM HEPES [pH 7], 150 mM KCl, 100 μg/ml BSA, 0.5 mM TCEP. Before all experiments, we determined the quality of lipid bilayer by charging beads with a mixture of 50 nM his 10 GFP and 50 nM his 10 Cherry-SCAR. If the beads are uniformly coated with GFP and Cherry, you can assume the beads have a continuous and uniform membrane. However, if you observed bright cherry fluorescent patches on glass beads that do not colocalize with GFP, a non-continuous bilayer was generated. We also observed that his 10 Cherry-SCAR, but not his 10 GFP bound to bare glass microspheres under these conditions.
For experiments in which we froze actin assembly and disassembly with the JLY cocktail, cells were allowed to spread on PLL coated glass for 30 min in the presence of 10 μM Rock kinase inhibitor (Y27632). Cells were then treated with 10 μM Jasplakinolide (CalBiochem, Cat# 420107), 8 μM Latrunculin B (Enzo Life Science, Cat# T110-0001), and 10 μM Y27632 Rock kinase inhibitor (CalBiochem, Cat# 688001) to freeze actin assembly and disassembly in XTC cells (Peng et al., 2011). All drugs were diluted to a 2× final concentration in 70% L-15 media lacking serum and penicillin/ streptomycin before being added to cells.
B16F1 mouse melanoma cells (ATCC CRL-6323) were cultured in Dulbecco Modified Eagles High glucose containing 10% heat inactivated fetal bovine serum and penicillin/streptomycin. Cells were grown in 25 cm 2 flasks at 37˚C in the presence of 5% CO 2 and split every 3-4 days. For transfection of pCMV-EGFP-Lpd (850-1250aa), cells were plated in 24-well dish to a confluency of 50-60%. We added 0.5-1.0 μg DNA in complex with 2.5-5.0 μl Superfect (Qiagen, Cat# 1006699) to each well containing complete media. After 5-6 hr, the media was changed. 24-36 hr after transfection, cells were prepared for live cell imaging.
Circular glass coverslips (25 mm, Warner Instruments Cat# 64-0715) were cleaned with 3 M NaOH for 30 min at 23˚C. Coverslips were rinsed extensively with MilliQ water and coated with 50 μg/ml mouse laminin (Sigma, Cat# 23017-015) diluted in PBS [pH 7.2] at 37˚C for 2 hr. Laminin coated coverslips were rinse with PBS and then assembled in stainless steel Attofluor cell imaging chamber (Invitrogen). Transfected cells were then trypsinized in 24-well plastic dishes, centrifuged, wash with complete media, and seeded on the laminin coated glass in the presence of filter sterilized Ham's F12 media containing 10% FBS and 50 mM HEPES [pH 7.2]. Cells spread on laminin coated glass within 30 min. Polarized cells were imaged using wide-field epifluorescence and a Nikon Plan Apo 60× TIRF (NA 1.45) objective on a Nikon Eclipse microscope.

Microscopy and image processing
XTC cell images were acquired on an inverted Nikon Eclipse TIRF microscope using either a 60× Nikon Plan Apo 60× TIRF (NA 1.45) or a 100× Nikon TIRF (1.49 NA) objective at 23-25˚C. B16F1 cells were imaged with Nikon Plan Apo 60× TIRF (NA 1.45) using wide-field epifluorescence illumination at 37˚C in the presence of CO 2 . We employed the Nikon Perfect Focus instrument to maintain the focal plan throughout image acquisition. GFP/Alexa488, Cy3, and Cy5 fluorophores were excited with 491 nm, 561 nm, and 638 nm Coherent lasers respectively. Laser power was modulated with neutral density filters, an AOTF, and exposure settings such that 1-2 mW of laser power was delivered through the objective as measured with a power meter. We used a single filter cube containing a multipass excitation and dichroic filter (Chroma). A rapid switching Sutter Instruments emission filter wheel was positioned before our camera. Images of XTC cells, B16F1 cells, and single actin filament TIRF assay were collected on a cooled Andor Xion EM-CCD camera. The microscope, camera, and lasers were controlled using Micromanager 1.4 (Edelstein et al., 2010). Image processing and data analysis was performed using ImageJ. Kymographs were generated using the ImageJ plugin, MultipleKymographs. Data was graphed using Kaleidagraph and Prism. Figures for the manuscript were made in Adobe Illustrator CS6.