The bacterial virulence factors VopL and VopF nucleate actin from the pointed end

Diverse actin filament networks are assembled, maintained, and disassembled by different sets of actin-binding proteins with complementary biochemical activities (Blanchoin et al., 2014). These networks are initiated by actin assembly factors that enhance the transition from monomeric to F-actin by catalyzing nucleation, elongation, or both. Better understood are the actin assembly mechanisms of Arp2/3 complex and formins (Chhabra and Higgs, 2007; Pollard, 2007; Chesarone and Goode, 2009; Campellone and Welch, 2010). Arp2/3 complex associates with the side of a mother filament and nucleates a new branch (daughter) filament from the pointed end. Formins make long, straight filaments by nucleating and remaining processively associated with elongating filament barbed ends.

In contrast, there is significant uncertainty and disagreement about the mechanisms of the third major class of actin assembly factors, tandem WH2 domain–based nucleators (Dominguez, 2016). Proteins in this group include bacterial (Rickettsia Sca2 and Vibrio VopL and VopF) and eukaryotic (Cordon-bleu [Cobl] and Spire) actin assembly factors, with different numbers and configurations of actin monomer–binding WH2 domains, proposed to promote filament nucleation and elongation by different mechanisms (Paunola et al., 2002; Dominguez, 2007, 2016; Qualmann and Kessels, 2009; Carlier et al., 2011). Sca2 contains three WH2 domains and promotes actin filament nucleation and elongation by remaining processively associated with the barbed end (Haglund et al., 2010; Madasu et al., 2013). Spire contains four WH2 domains but functions optimally as a dimer, induced by interaction with formins, and nucleates actin filament assembly from the pointed end (Quinlan et al., 2005, 2007; Rasson et al., 2015). Cobl contains a tandem repeat of three WH2 domains (Ahuja et al., 2007) and was originally identified as an actin nucleator involved in neuronal cell morphogenesis (Ahuja et al., 2007; Haag et al., 2012). However, other studies have produced conflicting models about the activities of Spire and Cobl, concluding that Spire functions mainly as a barbed-end capping protein (CP; Montaville et al., 2014), whereas Cobl is only a weak nucleator and functions primarily as a monomer-sequestering and filament-severing factor (Carlier et al., 2011; Husson et al., 2011).

Another example of disagreement concerns the filament assembly mechanism of VopL and VopF (VopL/F), which are virulence factors from the pathogenic gram-negative bacteria Vibrio parahaemolyticus and Vibrio cholera (Dominguez, 2016). VopL/F are type III secretion factors that contribute to entry and infection of host intestinal epithelial cells by facilitating nonphysiological actin cytoskeleton rearrangements (Reidl and Klose, 2002; Liverman et al., 2007; Tam et al., 2007, 2010). VopL/F are closely related (32% sequence identity and 72% sequence similarity), and both contain three WH2 domains and a ∼240-residue VopL/F C-terminal domain (VCD) that mediates dimerization (Fig. 1 A; Namgoong et al., 2011; Yu et al., 2011). VopL/F induce actin polymerization in cells and nucleate actin assembly in vitro (Liverman et al., 2007; Tam et al., 2007). However, despite their identical domain organization (Fig. 1 A) and similar sequences, studies have reached opposite conclusions regarding their mechanism of actin assembly (Namgoong et al., 2011; Yu et al., 2011; Pernier et al., 2013; Avvaru et al., 2015).

VopL has been proposed to nucleate actin filament assembly from the pointed end, whereby the WH2 domains and VCD cooperate to bind and nucleate actin monomers, leaving the VCD domain associated briefly with the pointed end of the nascent filament (Namgoong et al., 2011; Yu et al., 2011). A crystal structure of VopL in complex with actin revealed that the VCD dimer organizes three actin subunits into a filament-like conformation, with the VCD bound at the pointed end of this trimeric actin nucleus (Zahm et al., 2013). The WH2 domains are an accessory, contributing to nucleation by delivering actin subunits to the VCD-assembled nucleus (Namgoong et al., 2011; Yu et al., 2011; Zahm et al., 2013). Visualization by total internal reflection fluorescence microscopy (TIRFM) indicates that VopL associates briefly with the pointed end of the filaments it nucleates (Namgoong et al., 2011). However, ∼7.6% of the filaments associated with quantum dot–conjugated VopL buckled and elongated at twice the control rate, indicative of processive barbed end association. These results suggested a “template” nucleation mechanism whereby VopL stimulates rapid (shotgun) cycles of nucleation and associates only briefly with the pointed end of newly formed filaments through its VCD (Namgoong et al., 2011; Yu et al., 2011; Zahm et al., 2013).

In contrast, single-color TIRFM and bulk pyrene-actin assembly studies on VopF led to the conclusion that it sequesters actin monomers and associates exclusively with the barbed end to compete with CP and promote processive elongation (Pernier et al., 2013; Avvaru et al., 2015). Additionally, small-angle x-ray scattering analysis of progressively smaller VopF constructs in complex with actin suggested that VopF stabilizes linear strings of actin subunits, instead of a double-stranded filament (Avvaru et al., 2015). Combined, these results led to a model whereby VopF can associate with the side of a growing filament and potentially ride processively on the barbed end, adding monomers and preventing capping (Pernier et al., 2013).

Here, we sought to resolve this controversy by using multicolor TIRFM to directly observe spontaneous and seeded actin monomer assembly in the presence of fluorescently labeled VopL/F. The results allow us to conclude that VopL/F catalyze actin nucleation by identical pointed-end mechanisms and bind barbed ends only under nonphysiological experimental conditions.

We purified VopL/F constructs containing the two proline-rich domains, WH2 domains, and VCD (Fig. 1 A), and lacking only the variable N-terminal region, which is dispensable for actin assembly (Liverman et al., 2007; Namgoong et al., 2011; Pernier et al., 2013). A SNAP-tag was added at the N terminus of each construct for labeling with fluorescent dyes, hereafter referred collectively as SNAP-VopL/F, as the activities of both proteins were nearly indistinguishable in all the assays.

We first compared the activities of untagged, unlabeled, and fluorescently labeled versions of VopL/F by monitoring the assembly of Mg-ATP-actin (10% pyrene actin) in bulk pyrene-actin polymerization assays (Fig. S1). All the proteins accelerated spontaneous actin monomer assembly in a nearly identical way with increasing concentration (Figs. S1, A–C), demonstrating that the SNAP-tag and fluorescent labels do not affect their activities. Moreover, the time to half-maximal pyrene fluorescence agreed well with previous results using unlabeled VopL/F constructs (Liverman et al., 2007; Tam et al., 2007; Namgoong et al., 2011; Pernier et al., 2013).

We then used multicolor TIRFM to directly visualize actin assembly by fluorescently labeled SNAP-VopL/F. We monitored the assembly of 1.5 µM Mg-ATP-actin monomers (15% Oregon-green actin) with 0.2 nM red-labeled 549-SNAP-VopL/F (Fig. S1, E and F; and Video 1). 549-SNAP-VopL/F nucleate actin polymerization and remain bound to one end of the actin filament as it begins to elongate. Proteins that remain continuously associated with elongating barbed ends, such as formin (Kovar, 2006), typically modify the elongation rate. By comparison, filaments associated with 549-SNAP-VopL/F elongated at a rate of ∼13.0 subunits per second (sub s^–1), indistinguishable from the elongation rates of control filaments and filaments dissociating from 549-SNAP-VopL/F (Fig. 1, B and C). We also confirmed that VopL/F function as dimers, as reported (Namgoong et al., 2011; Pernier et al., 2013; Zahm et al., 2013), because 88% of the 549-SNAP-VopL/F molecules on nucleated filaments photobleached in two or fewer steps, as predicted based on a labeling efficiency of ∼75% (Figs. S1, G and H).

The mean time that 549-SNAP-VopL/F remained associated with nucleated filaments before dissociation was 27–34 s (Fig. 1 D, red and blue traces). However, these filaments are not visible immediately after nucleation because of the dead time required for the reaction to flow into the chamber (∼20 s) and for filaments to reach an observable length (∼0.5 µm). Therefore, assuming that VopL/F were bound to these filaments from their inception (nucleation), we determined their actual mean residence time to be ∼110 s (Fig. 1 D, orange and purple traces), during which time ∼1,400 actin subunits were added to the filaments. Higher oligomeric states of 549-SNAP-VopL/F do not correlate with increased residence times (Fig. S1 I). We conclude that in spontaneous actin monomer assembly reactions, individual VopL/F dimers display exclusively actin nucleation activity (not elongation or capping) and remain associated with one end of the filament for a short time of ∼110 s.

The pointed ends of actin filaments assembled from fluorescently labeled monomers bleach much faster than their barbed ends, to which fresh subunits add at a faster rate (Fig. S2 A; Kovar and Pollard, 2004). Thus, we imaged TIRFM reactions rapidly (1 frame/s) to observe spontaneous actin assembly with 549-SNAP-VopL/F (Fig. 1, E and F; and Video 2). Linescans of filament fluorescence revealed a steady decrease in intensity from the barbed to the pointed end (Fig. 1, G–J). VopL/F were always bound to the dimmer, pointed end of nucleated filaments (Fig. 1, H and I, n ≥ 101). Conversely, the formin mDia2 (549-SNAP-mDia2), a well-known barbed-end actin assembly factor, localized exclusively to the bright, barbed end of actin filaments (Figs. 1 J and S2 B; and Video 2; Kovar, 2006).

We can rule out the possibility that the SNAP-tag artificially induces VopL/F-mediated pointed-end nucleation, because fluorescently labeled CP (647-SNAP-CP) bound similarly to the barbed ends of filaments nucleated by untagged VopL and 549-SNAP-VopL (Fig. S1 D). Furthermore, C-terminally tagged VopL/F-SNAP-549 (Fig. S2, C–E) and 549-SNAP-VopL, at a range of concentrations (200–800 pM; Fig. S2 F), also nucleated exclusively from the pointed end, demonstrating that the position of the tag and protein concentration do not alter the nucleation behavior of VopL/F.

Given that VopL/F are pointed-end nucleators, the barbed-end binding proteins formin and CP should bind to the opposite end. We used three-color TIRFM to directly visualize the assembly of 1.5 µM Mg-ATP-actin (15% Oregon green actin) monomers with 549-SNAP-VopL/F and cyan-labeled CP (647-SNAP-CP) or mDia2 (647-SNAP-mDia2; Fig. 2). As expected, in 100% of our observations, 647-SNAP-mDia2 and 647-SNAP-CP associated with the opposite end of 549-SNAP-VopL/F-nucleated actin filaments (Fig. 2, A–E). Furthermore, 647-SNAP-mDia2 and 647-SNAP-CP never appeared to dissociate or replace 549-SNAP-VopL/F at the pointed end (not depicted).

In the presence of actin monomers, VopL/F bind exclusively to the pointed end of the filaments they nucleate (Figs. 1 and 2) and, unlike mDia2, were never observed to either bind the ends of spontaneously assembled control filaments or rebind to the ends of the filaments from which they had dissociated after nucleation. These observations contradict a previous study that VopF binds to both free and capped barbed ends (Pernier et al., 2013). To explore this question further, we directly tested the ability of VopL/F to bind preassembled filament ends (Fig. 3 and Video 3). Actin filaments were preassembled in a TIRFM chamber from 1.5 µM Mg-ATP-actin (10% Alexa Fluor 488 actin), and 549-SNAP-VopL/F (Fig. 3, A and C) or 549-SNAP-mDia2 (Fig. 3, B and D) was flowed in with or without 1.5 µM Mg-ATP-actin monomers (10% Alexa Fluor 488 actin). The presence of actin monomers resulted in a marked difference in the lengths of filaments bound by 549-SNAP-VopL/F compared with 549-SNAP-mDia2 (Fig. 3, A, B, and E; and Video 3). mDia2 associates with the barbed ends of the filaments it nucleates and to preassembled filaments (Kovar et al., 2006). 549-SNAP-mDia2 was observed bound to filaments with a large distribution of lengths, including shorter, nucleated filaments and longer, preassembled filaments (Fig. 3, B and E, mDia2 red dots). Consistent with our previous observations (Figs. 1 and 2), 549-SNAP-VopL/F exclusively associated with one end of shorter (∼1.5-µm-long) nucleated filaments (Fig. 3, A and E, VopL/F red dots).

In contrast, both 549-SNAP-VopL/F and 549-SNAP-mDia2 bound preassembled filaments ends when actin monomers were not included in the reaction (Fig. 3, C and D; and Video 4). 549-SNAP-mDia2 bound the barbed end of preassembled filaments with a broad length distribution (Fig. 3 E, mDia2 black dots), and in 117 of 118 observations, it remained bound throughout the 10-min experiment. Similarly, 549-SNAP-VopL/F (Fig. 3 C) can bind preassembled actin filament ends in the absence of actin monomers (Fig. 3 E, VopL/F black dots). However, VopL/F associate with both barbed and pointed ends, with no significant preference (Fig. 3, F and H), and with a mean residence time of ∼25–30 s (Fig. 3, G and H).

Because cells contain a mixture of actin filaments and monomers that are mostly bound to profilin, we modified the preassembled actin filament TIRFM assay. Green actin filaments were preassembled in a TIRFM chamber from 1.5 µM Mg-ATP-actin monomers (10% Alexa Fluor 488 actin). Red-labeled assembly factors (549-SNAP-VopL/F or 549-SNAP-mDia2) were then flowed into the chamber with 1.5 µM cyan Mg-ATP-actin monomers (10% Alexa Fluor 647 actin) in either the absence or presence of profilin at a 2:1 (3.0 µM) or 5:1 (7.5 µM) molar ratio to actin monomers (as described in Fig. 4 A). This setup reveals whether the assembly factors associate with cyan-colored newly nucleated filaments or green/cyan-colored preassembled filaments, whose barbed ends elongate from cyan-labeled actin (Fig. 4 A, right).

From such reactions (Fig. 4 B), we were able to quantify the following: (a) the percentage of nucleated or preassembled filaments associated with assembly factors (Fig. 4 C); (b) the lifetimes of association of assembly factors with nucleated and preassembled filament ends (Fig. 4 D), and (c) the barbed-end elongation rates of nucleated and preassembled filaments associated with assembly factors (Fig. 4 E). In the absence of profilin, VopL/F associated exclusively with the bleached pointed ends of the filaments they nucleated (Fig. 4 C) for ∼80 s (Fig. 4 D), which elongated at the control barbed end rate of ∼10 sub s^–1 (Fig. 4 E). In contrast, mDia2 associated with the barbed ends of both nucleated and preassembled filaments (Fig. 4 C), slowing the barbed end elongation rate to ∼2.5 sub s^–1 (Fig. 4 E; Kovar et al., 2006). In the presence of two- or fivefold molar excess of profilin to actin monomers (Fig. 4 B), the vast majority (85–95%) of VopL/F also associated with the pointed end of newly nucleated filaments (Fig. 4 C) for ∼80 s (Fig. 4 D), which elongated at ∼10 sub s^–1 (Fig. 4 E). However, in these reactions, 5–15% of the VopL/F association events were with the barbed ends of preassembled actin filaments (Fig. 4 C and Video 5) for ∼25 s (Fig. 4 D), in which case barbed-end elongation is essentially capped (Fig. 4 E). In contrast to flow-in experiments without actin monomers (Fig. 3), VopL/F were not observed to interact with the pointed end of preassembled actin filaments in the presence of profilin. Inclusion of profilin also slightly increased the percentage of mDia2 associated with preassembled filaments (Fig. 4 C), likely because of profilin’s inhibition of formin-mediated nucleation (Kovar et al., 2003). In conclusion, even if the presence of a large excess of profilin allows VopL/F to associate with and cap filament barbed ends (Pernier et al., 2013), the vast majority of observations (∼85%) correspond to nucleation events from the pointed end.

Previous studies reported opposite actin assembly mechanisms for the closely related Vibrio virulence factors VopL and VopF; VopL exclusively nucleated polymerization from the pointed end (Namgoong et al., 2011; Zahm et al., 2013), whereas VopF exclusively bound to the barbed end of preassembled actin filaments (Pernier et al., 2013; Avvaru et al., 2015). By addressing this discrepancy, we discovered here that VopL/F share a multifunctional mechanism of actin filament assembly and association (Fig. 5). The vast majority of filament assembly events, with or without saturating profilin, converge on a pointed-end nucleation mechanism for VopL and VopF (Namgoong et al., 2011; Yu et al., 2011; Zahm et al., 2013; Fig. 5, A and C). Furthermore, our experiments can reproduce all the filament association modes previously reported for VopL/F (Namgoong et al., 2011; Pernier et al., 2013; Zahm et al., 2013; Avvaru et al., 2015) and reveal that the reaction conditions critically determine whether VopL/F associate with different filament ends.

We propose that the primary mechanism of actin filament assembly by VopL/F is as follows. The WH2 domains recruit actin monomers to the VCD, which organizes three monomers into a nucleus (Zahm et al., 2013). VopL/F remain briefly bound to the pointed end via the VCD domain, whereas the WH2 domains dissociate to allow the nucleus to elongate from the barbed end (Fig. 5 A). After detachment of the WH2 domains, the affinity of VCD alone for the pointed end is relatively weak, which at least in part explains the dissociation of VopL/F after only ∼100 s. Upon dissociation, VCD might carry actin subunits with it, which can act as a new nucleation center, overcoming inhibition of nucleation by profilin; or alternatively, the VCD could reload profilin-free monomers from solution.

Unlike other F-actin end-binding proteins, such as formins and CP, VopL/F can bind the ends of preassembled filaments only under nonphysiological conditions, including in the total absence of actin monomers, where they can bind to both the barbed and pointed end (Fig. 5 B), or infrequently (<15% of events), when actin monomers are fully saturated with profilin, where they can bind only to the barbed end (Fig. 5 C). The VCD domain likely facilitates association of VopL/F with preassembled pointed ends (Fig. 5 B; Namgoong et al., 2011; Zahm et al., 2013), whereas the WH2 domains might mediate binding to the barbed end of preassembled filaments (Fig. 5, A and C). Indeed, although the WH2 domain preferentially binds actin monomers, if monomers are omitted from the reaction or the binding site of the WH2 domain on the actin monomer is masked by excess profilin, they might bind to the barbed end of the filament, which is the only location along the filament where the binding site of the WH2 domain is fully exposed (Chereau et al., 2005; Dominguez, 2010, 2016). Finally, why do VopL/F not bind to the pointed end of preassembled filaments in the presence of saturating profilin? Under these conditions, profilin and profilin-actin likely saturate the proline-rich domains of VopL/F, which may sterically hinder binding to the pointed end. Together, the preferential binding of VopL/F to actin monomers, their quick detachment from the pointed end of newly assembled filaments, and their negligible binding to preformed filaments converge to favor nucleation as their main activity under physiological conditions.

Our results explain some of the disagreement among previous studies, as some of the experiments were performed with a 5:1 profilin:actin ratio. However, it is unclear how the studies on VopF failed to observe pointed-end binding (Pernier et al., 2013; Avvaru et al., 2015). We found that pointed-end nucleation is the only mechanism for VopL/F in the presence of actin monomers, whereas in the absence of actin monomers, VopL/F bind to both filament ends, and even with saturating profilin, the vast majority of events (∼85%) correspond to pointed-end nucleation.

The cartoon representing our combined results of actin filament nucleation and association by VopL/F (Fig. 5) agrees well with in vivo studies (Liverman et al., 2007; Tam et al., 2007, 2010). The primary function of VopL/F is to promote virulence of the Vibrio bacteria through efficient pointed-end nucleation of unproductive actin filaments in the host cell (Fig. 5, A and C). Given that the concentration of actin monomers in Vibrio’s preferred hosts (higher eukaryotes) can be as high as 100 µM (Pollard et al., 2000), VopL/F will primarily drive efficient pointed-end nucleation. Furthermore, even if actin monomers are outnumbered 5:1 by profilin (Goldschmidt-Clermont et al., 1991, 1992; Wühr et al., 2014), pointed-end nucleation should predominate (Figs. 4 and 5 C). Assembly of unproductive F-actin decreases the concentration of actin monomers in host cells to the detriment of endogenous F-actin networks. Both in vitro electrophysiological and in vivo mouse colonization results indicate that VopF negatively affects F-actin–dependent tight-junction integrity in intestinal epithelial cells (Balda and Matter, 2008; Tam et al., 2010), which can lead to intestinal inflammation, rapid fluid loss, and passing of the Vibrio bacteria (Faruque et al., 1998; Reidl and Klose, 2002).

VopL/F constructs were prepared by infusion (Takara Bio Inc.) after PCR amplification (iProof; Bio-Rad Laboratories) from plasmid DNA for VopL and from codon-optimized plasmid DNA for VopF (provided by E. Kerkhoff, University of Regensburg, Regensburg, Germany), with a 6xHis tag included in the reverse primer. PCR products were cloned into the SNAP-tag-T7-2vector (New England Biolabs, Inc.) at the XmaI/PacI sites, but the PacI site is not maintained. A flexible linker (GGSGGS) was included in the forward primer sequences of the SNAP constructs between the SNAP and the VopL/F DNA sequences.

Recombinant SNAP-VopL/F-6xHIS, VopL/F-SNAP-6xHIS, SNAP-MmDia2 (521-1171, FH1-C)-6xHIS (SNAP-mDia2), and mouse CP (dual-expression construct of MmCP α1-6xHIS and SNAP-β2; SNAP-CP) proteins were purified from Escherichia coli strain BL21-Codon Plus (DE3)-RP (Agilent Technologies), after their expression with 0.5 mM isopropyl β-d-1-thio-galactopyranoside for 16 h at 16°C. Cells were lysed with an Emulsi-Flex-C3 (Avestin) in extraction buffer (50 mM NaH₂ PO₄, pH 8.0, 500 mM NaCl, 10% glycerol, 10 mM imidazole, and 10 mM β-mercaptoethanol) and were clarified by centrifugation at 16,000 rpm for 15 min, followed by 18,000 rpm for 30 min, in a Beckman SS-34 rotor. The extract was incubated for 1 h at 4°C with Talon Resin (Takara Bio Inc.), loaded onto a column, and washed with extraction buffer, and all proteins were eluted with 250 mM imidazole. SNAP-VopL/F and SNAP-mDia2 were dialyzed against buffer (20 mM Hepes, pH 7.4, 200 mM KCl, 0.01% NaN₃, and 1 mM DTT). SNAP-CP was dialyzed against CP buffer (10 mM Tris, pH 7.5, 40 mM KCl, 50% glycerol, 0.5 mM DTT, and 0.01% NaN₃). SNAP-tagged VopL/F were labeled with SNAP-Surface 549, and SNAP-mDia2 and SNAP-CP were labeled with SNAP Surface 549 or 647 according to the manufacturers’ protocols (New England Biolabs, Inc.). Concentrations of SNAP-tagged proteins and the degree of labeling were determined by densitometry of Coomassie-stained bands on SDS-PAGE gels compared with standards and by measuring fluorophore absorbance in solution using the extinction coefficient of SNAP-surface549: ε₅₆₀ = 150,000 M/cm. VopL/F were flash-frozen in liquid nitrogen and stored at −80°C. Actin was purified from chicken skeletal muscle acetone powder by a cycle of polymerization and depolymerization and gel filtration (Spudich and Watt, 1971). Gel-filtered actin was labeled on Cys374 with Oregon green iodoacetamide (Thermo Fisher Scientific; Kuhn and Pollard, 2005) or on surface lysines with Alexa Fluor 488 or Alexa Fluor 647 succinimidylester (Thermo Fisher Scientific; McCullough et al., 2011; Kang et al., 2012). Additional information about protein purification for mouse CP can be found in Palmgren et al. (2001), and for formin mDia2, in Suarez et al. (2015).

Microscope slides and coverslips (#1.5; Thermo Fisher Scientific) were washed successively for 10 min each with (1) acetone, (2) isopropanol, (3) MilliQ H₂O, and (4) isopropanol. They were then sonicated for 30 min with isopropanol, incubated for 1 h with piranha solution (80% H₂SO₄ and 20% H₂O₂), washed with deionized water, then washed with isopropanol and dried. Glass was incubated for 18 h with 1 mg/ml mPeg-Silane (5,000 MW) in 95% ethanol, pH 2.0. Slides were washed once with 95% ethanol and then five times with MilliQ H₂O, air dried, and stored in Petri dishes at 4°C. Parallel strips of double-sided tape were placed on the coverslip to create multiple flow chambers (Zimmermann et al., 2016).

TIRFM images were collected at 1- to 4-s intervals with an iXon EMCCD camera (Andor Technology) using an Olympus IX-71 microscope fitted with through-the-objective TIRFM illumination with an Olympus 60× 1.49-NA oil objective controlled by MetaMorph Basic Version 7.8.2.0 (Molecular Devices) at 22°C (Zimmermann et al., 2016). Mg-ATP-actin (15% Oregon green actin) was mixed with 2× TIRF buffer (1× = 10 mM imidazole, pH 7.0, 50 mM KCl, 1 mM MgCl₂, 1 mM EGTA, 50 mM DTT, 0.2 mM ATP, 50 µM CaCl₂, 15 mM glucose, 20 µg/ml catalase, 100 µg/ml glucose oxidase, and 0.5% [400 centipoise] methylcellulose) and SNAP-VopL/F and transferred to a flow cell for imaging at 23°C. For two- and three-color TIRFM, we cyclically imaged Oregon green actin or Alexa Fluor 488 actin (one frame, 488-nm excitation for 50 ms), 549-SNAP-VopL/F (one frame, 561-nm excitation for 50 ms), and 647-SNAP-CP, 647-SNAP-mDia2, or Alexa Fluor 647 actin (one frame, 640-nm excitation for 50 ms).

For photobleaching experiments to determine the oligomerization state of VopL/F, 549-SNAP-VopL/F were imaged continuously, whereas actin filaments were imaged intermittently (Winkelman et al., 2014). Specifically, TIRFM images were collected continuously (10–20 frames/s) in the 561-nm channel, with the 488-nm shutter opened every 20–40 frames to visualize actin. The observed distribution of photobleaching steps was compared with the number of steps predicted by a binomial distribution model where s = 0.75 (Breitsprecher et al., 2012). When spots did not bleach completely, the number of fluorescent subunits present was inferred by intensity and step size. Background fluorescence was subtracted with ImageJ using a rolling ball radius of 30 pixels, and the integrated fluorescence signals of 549-SNAP-VopL/F spots associated with the end of actin filaments were manually tracked and measured using ImageJ.

To calculate residence times on actin filament ends in all conditions, 549-SNAP-VopL/F spots associated with filament ends were tracked manually in ImageJ (Winkelman et al., 2014). Residence times for 549-SNAP-VopL/F on nucleated actin filaments in spontaneous TIRFM assays in Figs. 1 D and 4 D were determined through back-calculation by measuring the length of actin filaments immediately before the 549-SNAP-VopL/F dissociated and converting that length into total actin subunits (1 µm = 375 subunits). The subunit length was divided by the mean elongation rate of the filaments. Barbed-end elongation rates were calculated by measuring filament lengths over time with ImageJ. Lengths of actin filaments for the dot plot in Fig. 3 E were measured in ImageJ immediately upon binding by 549-SNAP-VopL/F or 549-SNAP-mDia2. We deemed an event to be bound if 549-SNAP-VopL/F or 549-SNAP-mDia2 colocalized with the end of a filament for at least two frames.

Partially bleached actin filaments were quantified by using the linescan function in ImageJ. A line was drawn down the length of the filament at the last frame in which a 549-SNAP-VopL/F or 549-SNAP-mDia2 was associated in the RGB merged image and saved as an region of interest (ROI). These ROIs were imported into the 488-channel time lapse (with background subtracted), and a linescan was taken at this point. The raw values were normalized from 0 to 1 for each individual filament. To account for noise, the mean was taken of the three normalized values proximal or distal to the 549-SNAP-VopL/F or 549-SNAP-mDia2. If the mean of the normalized values was less than 0.2 it was considered to be in the dimmest 20% of the filament, and if the value was higher than 0.8 it was considered to be in the brightest 20%. All 549-SNAP-VopL/F events were bound to the dimmest 20%, whereas all 549-SNAP-mDia2 events were bound to the brightest 20%.

Bulk actin assembly was measured from the fluorescence of pyrene-actin with a Safire2 (Tecan) fluorescent plate reader (Neidt et al., 2008). In brief, the assembly of 10% pyrene-labeled Mg-ATP-actin monomers was initiated by the addition of 50 mM KCl, 1 mM MgCl₂, 1 mM EGTA, 10 mM imidazole, pH 7.0, and other proteins to be assayed. Final protein concentrations are indicated in the figure legends.

Fig. S1 shows that the SNAP-tag does not affect VopL/F’s activities via bulk pyrene-actin assembly assays and three-color TIRFM as well as the photobleaching of 549-SNAP-VopL/F molecules from in vitro TIRFM. Fig. S2 shows that VopL/F-SNAP-549 (C-terminally tagged constructs) localize to the dim, pointed end of partially bleached fluorescently labeled actin filaments. In Videos 1–5, all actin filaments are fluorescently labeled and visualized with multicolor TIRFM. Video 1 (related to Fig. 1) shows 549-SNAP-VopL/F localize to one end of actin filaments. Video 2 (related to Fig. 1 and Fig. S2) shows that 549-SNAP-VopL/F are on the dim, pointed end of partially bleached actin filaments versus the formin 549-SNAP-mDia2, which localizes to the bright, barbed end. Video 3 (related to Fig. 3) shows that 549-SNAP-VopL nucleates only short actin filaments, compared with the formin 549-SNAP-mDia2, which can both nucleate and bind long preassembled actin filaments in the presence of actin monomers. Video 4 (related to Fig. 3) shows that 549-SNAP-VopL binds to both ends of preassembled actin filaments in the absence of actin monomers. Video 5 (related to Fig. 4) shows that 549-SNAP-VopF primarily nucleates actin filaments in the presence of saturating profilin, but also occasionally binds to the barbed end of preassembled actin filaments.

We thank Dr. Eugen Kerkhoff for the VopF cDNA, Michael Rosen and members of his laboratory for helpful comments, and members of the Kovar (Jonathan Winkelman, Cristian Suarez, and Dennis Zimmermann) and Dominguez (Malgorzata Boczkowska) laboratories for helpful discussions and technical assistance.

This work was supported by National Institutes of Health (R01 GM079265 to D.R. Kovar, MCB Training Grant T32 GM007183-39 to T.A. Burke, and R01 GM073791 to R. Dominguez).

The authors declare no competing financial interests.

Author Contributions: D.R. Kovar, R. Dominguez, and T.A. Burke designed the experiments. T.A. Burke and A.J. Harker biochemically purified proteins. T.A. Burke performed all experiments. T.A. Burke and A.J. Harker analyzed data. T.A. Burke, D.R. Kovar, and R. Dominguez wrote the manuscript. All authors contributed to the discussion and evaluation of the manuscript.