Actin-dependent propulsion of Listeria monocytogenes is thought to require frequent nucleation of actin polymerization by the Arp2/3 complex. We demonstrate that L. monocytogenes motility can be separated into an Arp2/3-dependent nucleation phase and an Arp2/3-independent elongation phase. Elongation-based propulsion requires a unique set of biochemical factors in addition to those required for Arp2/3-dependent motility. We isolated fascin from brain extracts as the only soluble factor required in addition to actin during the elongation phase for this type of movement. The nucleation reaction assembles a comet tail of branched actin filaments directly behind the bacterium. The elongation-based reaction generates a hollow cylinder of parallel bundles that attach along the sides of the bacterium. Bacteria move faster in the elongation reaction than in the presence of Arp2/3, and the rate is limited by the concentration of G-actin. The biochemical and structural differences between the two motility reactions imply that each operates through distinct biochemical and biophysical mechanisms.
The actin cytoskeleton is a system of proteins that can perform mechanical work. In some situations, myosin motors generate the force to perform work while actin filaments provide directional tracks for myosin translocation. In others, actin polymerization itself appears to perform mechanical work. New actin polymer can be generated by creating a filament de novo from G-actin subunits in a process referred to as nucleation. Alternatively, new polymer can be generated by extending existing filaments, which we refer to as elongation.
The frequency of nucleation is an important factor in determining the morphology of actin-based cellular structures (Svitkina and Borisy, 1999; Svitkina et al., 2003; Vignjevic et al., 2003) It may also influence the force producing mechanisms entailed in these structures. For example, nucleation is frequent in protruding sheets termed lamellipodia (Theriot and Mitchison, 1991) where actin filaments are typically short and organized in dendritic branches (Svitkina and Borisy, 1999). Nucleation is less frequent in protruding cylinders such as filopodia where new polymer may be formed solely through elongation of existing filaments (Mallavarapu and Mitchison, 1999), and in these structures filaments are organized into parallel bundles (Lewis and Bridgman, 1992; Svitkina et al., 2003). Although protrusion of lamellipodia (Theriot and Mitchison, 1991) and filopodia (Mallavarapu and Mitchison, 1999) are tightly coupled to actin polymerization, the distinct organization and generation of actin filaments in each structure raises the possibility that each might use a different mechanism to produce mechanical force.
Progress on the mechanism by which actin polymerization can perform mechanical work has been facilitated by the discovery that a number of intracellular pathogens, including Listeria monocytogenes, propel themselves through the host cytoplasm by assembling an actin comet tail that grows at the tail–bacterial interface (Cameron et al., 2000; Portnoy et al., 2002). This motility mechanism is probably similar to rocketing of endocytic vesicles (Taunton et al., 2000), and shares biochemistry with leading edge structures (Pollard and Borisy, 2003). Like lamellipodial protrusion, L. monocytogenes motility is thought to be driven by frequent, Arp2/3-dependent nucleation that generates a tail containing short, highly branched filaments (Tilney and Portnoy, 1989; Theriot et al., 1992). Reconstitution of L. monocytogenes motility in a biochemically defined system (Loisel et al., 1999) has permitted detailed investigation of the biochemical and biophysical mechanism coupling polymerization with frequent nucleation to mechanical force production (Bernheim-Groswasser et al., 2002; Wiesner et al., 2003). We currently lack a similarly tractable reconstituted system for dissecting the mechanism for force production by elongation-dominated structures like filopodia. Here, we show that L. monocytogenes can move in the absence of frequent nucleation and begin the biochemical and structural characterization of this form of motility.
L. monocytogenes motility initiates from an actin cloud whose formation requires actin nucleation by the Arp2/3 complex (Welch et al., 1997; May et al., 1999; Yarar et al., 1999). To test if motility continues to require Arp2/3 activity after cloud nucleation, motility assays were performed in perfusion chambers. Motility could then be dissected into two steps: tail nucleation and tail elongation. We used actin tagged with different fluorophores to reveal the step during which a given segment of the comet tail polymerized. Motility was initiated with cytosol containing Arp2/3 and green actin. This solution was then replaced with cytosol containing an Arp2/3 inhibitor and red actin. We used the cofilin homology and acidic region of N-WASP (CA) domain of N-WASP to inhibit Arp2/3. This domain binds Arp2/3 and competitively inhibits binding of Arp2/3 activators including the ActA protein of L. monocytogenes (May et al., 1999). Bacteria mixed with brain cytosol formed comet tails (Fig. 1 A). Addition of CA to cytosol inhibited all actin polymerization around the bacteria (Fig. 1 B). However, once Arp2/3 initiated motility, its activity was no longer required for comet tail assembly. Motility in the absence of Arp2/3-driven nucleation was evident from an additional segment of comet tail containing red actin after the initiating cytosol was replaced with cytosol containing CA (Fig. 1 C). The segment of tail formed by elongation in cytosol, CA, and red actin appeared thinner, stiffer, and less fuzzy than the initial tail formed with active Arp2/3. Replacing the initiating cytosol with CA and actin alone resulted in only a short stretch of newly assembled actin (Fig. 1 D). Therefore, after initiation by Arp2/3, comet tails continue to assemble in its absence in a reaction that requires cytosolic factors in addition to actin.
Actin filament arrays assembled by Arp2/3-dependent nucleation are characterized by their highly branched, dendritic organization (Mullins et al., 1998; Svitkina and Borisy, 1999; Cameron et al., 2001). We used EM to visualize actin filament organization in comet tails assembled in the presence or absence of Arp2/3-nucleating activity (Fig. 2). For comparison, we first looked at the structure of L. monocytogenes comet tails in infected rat kangaroo (PtK) cells (Fig. 2, A and B). Comet tails assembled in vivo appeared to consist of two populations of actin filaments. A dendritic network of highly branched filaments is ensheathed by a cylinder of bundles running parallel to the direction of movement. Comet tails assembled by L. monocytogenes moving in brain extracts fell into two different classes. One class contained both filament organizations with prominent bundles surrounding a branched interior network (Fig. 2 C). The other class consisted almost exclusively of the branched filaments (Fig. 2 D). The type of comet tail that forms is extract dependent. Of six different brain extracts, two extracts primarily formed the hybrid tail consisting of branched as well as parallel bundles of filaments, as shown in Fig. 2 C (24/28 comet tails examined contained both filament organizations, whereas the remaining four comet tails contained only branched filaments). The other four extracts formed comet tails consisting primarily of branched filaments, as shown in Fig. 2 D (67/67 comet tails examined). We do not know the source of the variability between cytosolic extracts. In the two-step reaction, we observed an abrupt structural change between the portion of the tail generated in cytosol alone and that generated in cytosol + CA (Fig. 2, E and F). This structural change was independent of the type of comet tail formed in step 1. The first part contains branched filaments, whereas the second part consists only of long bundles running parallel to the direction of movement. The bundles are still arranged into a cylinder, but now the interior is largely hollow. If the first portion of the comet tail consisted of both branched filaments and parallel bundles, the bundles in the second portion appeared to extend off the original bundles (Fig. 2 E; 11/11 examined). If the first portion of the comet tail consisted solely of branched filaments, they were stitched together into parallel bundles in step 2 (Fig. 2 F; 23/23 examined) in a manner analogous to bundle formation from a dendritic network described by Vignjevic et al. (2003). This structural transition within comet tails is also quite similar to that seen after addition of G-actin to detergent-extracted infected cells (Tilney et al., 1992), suggesting that our assay might be reporting on the same activity as that seen in the Tilney work. From our EM, we conclude that continued assembly of the branched network in the center of the tail requires Arp2/3 nucleation, whereas the sheath of parallel bundles can apparently elongate independent of frequent nucleation by Arp2/3.
Arp2/3-dependent L. monocytogenes motility has been reported with seven pure proteins (Loisel et al., 1999). We wished to identify the biochemical factors required for Arp2/3-independent motility. We were unable to reconstitute Arp2/3-independent motility using mixtures of known factors, and a direct fractionation approach quickly resulted in the loss of all activity even when fractions were mixed with known factors. This suggested that Arp2/3-independent motility requires more than one new factor. Therefore, to facilitate its biochemical dissection, we divided L. monocytogenes motility into a set of subreactions in an attempt to make at least one subreaction dependent upon a single factor. We divided L. monocytogenes propulsion into three steps (Fig. 3 A). In the first step, L. monocytogenes are preincubated in cytosol under conditions that prevent actin polymerization. This allows recruitment of essential factors to the bacterial surface. The second step initiates actin polymerization and motility in the presence of purified Arp2/3, capping protein, and actin. In the third step, the initiating mixture is replaced with actin, CA, and brain fractions to be tested for elongation activity. All three steps are required for Arp2/3-independent propulsion in step 3, and actin alone in step 3 was not sufficient for sustained propulsion.
We fractionated brain extracts to identify the additional factor required in step 3 for persistent motility. A purification scheme from brain cytosol ended with the isolation of a single polypeptide of 56 kD that was identified by mass spectrometry as the actin-bundling protein fascin (Fig. 3 B). Actin + CA + recombinant fascin expressed in bacteria and purified was sufficient, in step 3, to support continued comet tail assembly in the absence of Arp2/3 (Fig. 3 C). We conclude that fascin is the only soluble protein required in addition to actin for Arp2/3-independent motility.
The actin-bundling proteins fimbrin (Kocks and Cossart, 1993) and α-actinin (Dabiri et al., 1990) have been localized to actin comet tails of L. monocytogenes. To see if fascin localized to comet tails, L. monocytogenes-infected BSC-1 cells and XTC cells were stained for fascin (Fig. 4). Fascin localized to comet tails in both cell types.
We tested other bundling proteins for their ability to substitute for fascin in Arp2/3-independent motility. Reactions were performed using the three-step protocol described above, but using either fascin, fimbrin, α-actinin, or filamin in step 3 (the elongation step). All of these bundling proteins scored in the reaction when present at 1 μM (Fig. 5). For each bundling protein, ∼30–40% of the bacteria that assembled clouds or comet tails in step 2 elongated in step 3 (for fascin, 68/210; fimbrin, 54/177; α-actinin, 36/91; filamin, 24/83). Therefore, Arp2/3-independent elongation requires bundling of actin filaments, but is not a unique property of fascin.
To gain insight into the biophysical mechanisms involved in propulsion, comet tails assembled under Arp2/3-dependent and -independent conditions were visualized by thin-section EM. Incubating bacteria in Arp2/3 and actin alone resulted in the formation of actin clouds containing highly branched filaments (Fig. 6 A). Bacteria assembled comet tails after preincubation in cytosol in the presence of actin, Arp2/3, and capping protein. Filaments under these conditions are short and highly branched with no bundles or long filaments visible (Fig. 6 B). Filament organization under these conditions matches that seen in the presence of Arp2/3 and actin alone, the main difference being the assembly of a comet tail as opposed to a symmetric cloud. This demonstrates that L. monocytogenes moving under these conditions undergo frequent rounds of Arp2/3-mediated nucleation. When the nucleation mix is replaced with fascin, CA, and actin, the organization of the comet tail switches abruptly to bundles of filaments parallel to the direction movement (Fig. 6 C). Higher magnification views of the nucleation-based reaction show a high concentration of branched filaments directly behind the bacteria (Fig. 6 D). In contrast, in the elongation reaction filaments make extensive lateral contacts with the bacterial surface, but very few filaments are localized behind it (Fig. 6 E).
Because Arp2/3 has previously been considered central to L. monocytogenes propulsion, we took several steps to test if bundling-mediated elongation is truly independent of nucleation by Arp2/3. We performed a dose–response analysis of CA inhibition for the nucleation phase (step 2) and elongation phase (step 3) of our reaction (Fig. 7 A). The Arp2/3-dependent nucleation phase was sensitive to CA inhibition with an IC50 of ∼0.5 μM, whereas CA did not inhibit the fascin-dependent elongation step at concentrations tested up to 15 μM. Second, we tested the effect of a nonhydrolyzable ATP analogue, adenosine 5′-(β,γ-imido)triphosphate (AMPPNP; Fig. 7 B). AMPPNP inhibits nucleation by the Arp2/3 complex by blocking an activation step, but has no effect on the association of Arp2/3 with ActA (Dayel et al., 2001). In that experiment, AMPPNP blocked Arp2/3-dependent nucleation when present at 50-fold molar excess of AMPPNP over ATP, whereas actin polymerization in the first 2 min was not affected by the nucleotide in solution. After that, ATP bound to actin exchanges with nucleotide in solution and further polymerization is inhibited (Selden et al., 1999). In the Listeria assay, AMPPNP inhibited cloud formation with an IC50 of ∼0.75 μM which, under these assay conditions, is in 30-fold molar excess of ATP in solution in the nucleation step and 60-fold excess in the elongation step. In the three-step assay, we normally can readily detect cloud formation within the first 60–90 s after Arp2/3 and actin are introduced into the chamber in step 2. Therefore, inhibition of cloud formation by AMPPNP is likely due to inhibition of Arp2/3 activation. In contrast, elongation was insensitive to AMPPNP up to 4 mM. Each of these results, as well as the absence of branched filaments in the fascin-mediated reaction, argue that the fascin-mediated elongation phase, once initiated, is independent of Arp2/3-nucleating activity. To test if Arp2/3 might play a role other than nucleation in the elongation phase, we introduced Arp2/3 labeled with a green fluorophore in step 2 and assessed its localization after elongation with fascin in step 3 (Fig. 7 C). Arp2/3 localized to the portion of the comet tail assembled in the nucleation phase (step 2), but was not detected in the tail extension generated in the presence of fascin and CA in step 3 nor at the bacterial surface.
In all of the previous experiments, motility was inferred from morphology by the appearance of a long segment of red actin extending away from the green tail assembled in the presence of Arp2/3. To directly measure motility, we used time-lapse phase-contrast or differential interference contrast imaging in the same three-step assay in perfusion chambers. When bacteria are preincubated in cytosol and step 3 contains fascin and actin, the bacteria move at nearly constant rates (∼2 μm/min) for 6 min before they start to decelerate (Fig. 8, A and B). The deceleration is due to depletion of actin or fascin from solution because refreshing the perfusion mixture promoted a new round of elongation on a subset of the bacteria (unpublished data). In contrast, when step 3 contains actin alone, the bacteria move only a short distance (to a maximum of 4 μm) before they stop (Fig. 8, A and B). In either the presence of fascin + actin or actin alone, bacteria rarely detached from the comet tail. From these experiments, we conclude that an actin cross-linking protein is required for persistent movement in the absence of Arp2/3, but actin alone is sufficient for a short burst of motility.
To further characterize the nucleation- and elongation-driven reactions, we compared the rates of L. monocytogenes motility of all our different reactions as a function of G-actin concentration (Fig. 9). First, we tested motility in the presence of Arp2/3, VASP, capping protein, and actin in the absence of a cytosol preincubation. This is similar to the defined system for L. monocytogenes motility (Loisel et al., 1999), but it lacks profilin and cofilin so the F-actin is not depolymerizing or recycling, and the concentration of G-actin is not buffered. Under these conditions, L. monocytogenes move at a rate of 0.5–1 μm/min regardless of the G-actin concentration (1–16 μM). When this mixture is replaced with fascin and actin, all movement stops. If, instead, the bacteria are preincubated with cytosol and then mixed with Arp2/3, capping protein, and actin (equivalent to step 2 in our three-step reaction), they move faster than in the defined system. The rate depends on actin concentration at low G-actin concentrations and plateaus above 4 μM. L. monocytogenes move faster still when this mixture is replaced with fascin and actin (equivalent to step 3 in the three-step reaction), and the rate depends on the actin concentration at all of the concentrations tested. Thus, elongation-based motility is faster than Arp2/3-dependent motility. The elongation reaction absolutely requires one or more unknown factors that can bind to the bacterial surface, and these factors facilitate Arp2/3-dependent motility.
L. monocytogenes motility can be experimentally separated into an Arp2/3-dependent nucleation phase and an Arp2/3-independent elongation phase. Elongation-mediated propulsion, unlike propulsion mediated by frequent nucleation, requires an actin-bundling protein and an unknown factor that associates with the bacterial surface. Each biochemical mechanism generates a distinct comet tail structure. The additional biochemical requirements, alternative filament organization, and different dependence of rate on G-actin concentration all suggest that elongation-driven motility is not a simple continuation of the Arp2/3-mediated reaction. We surmise that the mechanical mechanisms for generating force likely differ in the two reactions.
Although the biophysical mechanism by which actin polymerization drives propulsion is not known for Arp2/3-dependent motility, it has been subject to extensive theoretical analysis (Merz and Higgs, 2003; Mogilner and Oster, 2003b; Upadhyaya and van Oudenaarden, 2003). Growing barbed ends near the bacterial surface are thought to either push or squeeze the bacteria forward or prevent it from diffusing backward. Comet tails generated by frequent Arp2/3-dependent nucleation contain a number of filaments directly behind the bacteria that are properly positioned to push as they grow. It is not clear whether these models can account for force production in the elongation-mediated reaction where very few filaments are localized behind bacteria. During elongation, the vast majority of filaments are bundled and bound to the sides of the bacteria. Despite the fact that these filaments are not positioned to push or prevent backward diffusion in the manner predicted by the elastic Brownian ratchet (Mogilner and Oster, 2003a), bacteria in fact move faster under elongation conditions than nucleation conditions. One reason for faster motility during elongation might be the absence of friction between the comet tail and the bacteria that normally limits the rate of motility in the presence of Arp2/3. Deformation of liposomes coated with ActA revealed a pulling force at the rear that retards forward movement (Giardini et al., 2003). Under elongation-driven conditions there are no filaments behind the bacteria, so this pulling force may not exist. Thus, actin polymerization becomes rate limiting.
The implication of the structural data is that the elongation reaction generates mechanical force using filaments bound to the bacterial surface along the side of the filament. If a myosin motor were attached to the surface of the bacteria, it could use ATP hydrolysis to walk up the laterally attached bundles. However, the elongation reaction is resistant to high concentrations of the nonhydrolyzable ATP analogue AMPPNP. In addition, we have not detected a myosin using a pan-myosin antibody in fractions enriched in the Listeria-bound factor, which is required for elongation, on Western blots. Therefore, we consider a myosin-based transport mechanism for L. monocytogenes propulsion highly unlikely even for our Arp2/3-independent elongation reaction.
The strong requirement for a bundling protein provides clues as to the biophysical mechanism of the elongation reaction. One possible function of fascin is simply to establish a stable base such that force generated by actin polymerization can be used for useful work. In the absence of a bundling protein, filaments longer than 30–150 nm would simply buckle under a load, and thus could not generate force in the Mogilner and Oster Brownian ratchet model (Mogilner and Oster, 1996). Arp2/3 repeatedly supplies a new population of branched, short filaments that are capable of exerting force which might explain the lack of a requirement for a bundling protein when it is present.
An alternative interpretation of the fascin requirement is that bundling plays a more direct role in force production. Bundling the filaments comprising the hollow cylinder assembled by the elongation reaction could squeeze the bacteria forward. Compressive forces are considered to drive motility in an elastic model of L. monocytogenes propulsion (Gerbal et al., 2000), and such forces have been shown to be exerted on ActA-coated liposomes (Giardini et al., 2003; Upadhyaya et al., 2003). We would favor such a mechanism if the comet tail assembled by the elongation showed any sign of collapsing behind the bacteria, but this is not seen. As an alternative, we consider a Brownian ratchet in which diffusion is rectified by bundling (Fig. 10). In this model, we hypothesize that individual filaments form weak attachments to the bacterial surface, but bundling of filaments is incompatible with attachment to the surface. Therefore, as bundling proceeds to form energetically favorable bonds, the bacterium diffuses forward to form favorable bonds with individual filaments growing in front of it. The total number of bonds in the system is maximized by forward movement of the bacterium. We have found that other actin cross-linking proteins such as fimbrin, α-actinin, and filamin can substitute for fascin in the elongation reaction. Perhaps comparing quantitative differences between the different bundling factors to formal models of force generation, in addition to identification of the unknown factor, can help distinguish the mechanism of elongation-driven motility.
Beyond its biophysical interest, is the elongation reaction physiologically significant for L. monocytogenes biology? Our EM of L. monocytogenes in infected cells and crude extracts suggests that physiological comet tails are typically surrounded by a sheath of bundled actin filaments. Polarization microscopy observations (Zhukarev et al., 1995) and EM of L. monocytogenes in protrusions (Sechi et al., 1997) are consistent with the presence of such a sheath. Our three-step assay suggests that this sheath is generated by the elongation reaction, whereas the gel inside the sheath is generated by Arp2/3-driven nucleation. The sheath might provide extra force for propulsion or may simply stiffen the comet tail. The morphology of comet tails we observed in the Arp2/3-driven and elongation reactions shows that the persistence length of the fascin-bundled sheath is much longer than the Arp2/3 cross-linked, branched gel, as expected from their structures. In infected cells, L. monocytogenes motility ceases and protrusions withdraw when the cells are injected with fragments of α-actinin (Dold et al., 1994). Therefore, comet tail stiffness might be important when L. monocytogenes needs to deform the plasma membrane in a filopodium-like protrusion, where the stiffness of the plasma membrane and submembranous cortex is opposing the stiffness of the comet tail. Unidirectional persistence of motility might also be important for the infectious life cycle because the efficiency of bacterial invasion between cells has been shown to correlate more with the unidirectional persistence of bacterial motility than with motility rate (Monack and Theriot, 2001). Testing the physiological significance of the elongation reaction will require depletion experiments in cells. Depletion of all bundling proteins in cells will be difficult, so the best approach may be to identify the missing Listeria-bound factor(s) that are required for the elongation reaction, and to test the effect of depleting it on L. monocytogenes biology. Identification of this factor(s) would also help elucidate the physiological process L. monocytogenes is hijacking to perform the elongation reaction.
Materials And Methods
Actin (Pardee and Spudich, 1982) and CapZ (Casella and Cooper, 1991) were purified from rabbit skeletal muscle as described. Arp2/3 was purified from bovine brain as described previously (Egile et al., 1999). Recombinant, 6His-tagged VASP (a gift from Frederick Southwick, University from Florida College of Medicine, Gainesville, FL) expressed in Sf9 cells with baculovirus was purified on nickel chelating resin followed by gel filtration on a Superdex™ 200 column. Recombinant fascin (a gift from Pierre McCrea, M.D. Anderson Cancer Center, Houston, TX) was purified as a thioredoxin fusion protein from bacteria (Tao et al., 1996) by conventional chromatography. Bacterial extracts were prepared in 20 mM Pipes, pH 6.8, 50 mM KCl, and 5 mM β-mercaptoethanol (β-Me). The extract was centrifuged for 1 h at 108,000 g. The supernatant was passed through a 2× 5-ml DEAE HiTrap™ column linked in tandem that flowed into a 5-ml S HiTrap™ column (Amersham Biosciences). The flow-through from these linked columns was dialyzed against 20 mM MES, pH 6.0, and 5 mM β-Me, applied to a Mono S 5/5 column, and eluted with a 20-ml gradient to 300 mM KCl in the dialyzing buffer. Fascin–thioredoxin eluted around 75–120 mM KCl. Fascin-positive fractions were gel filtered on a 16/60 Superdex™ 200 column equilibrated in 20 mM Pipes, pH 6.8, 100 mM KCl, and 5 mM β-Me. Recombinant fimbrin was a gift from Paul Matsudaira (MIT, Cambridge, MA). To purify recombinant fimbrin from bacteria, lysates were prepared in 20 mM Tris, pH 8.0, and 5 mM β-Me. The 108,000 g supernatant was precipitated with 30–70% ammonium sulfate and the pellet was resuspended in 20 mM Tris, pH 8.0, 50 mM KCl, and 5 mM β-Me. After dialyzing against the same buffer, the sample was applied to a 2× 5-ml DEAE HiTrap™ column. The column was eluted with a 15-volume gradient to 500 mM KCl in the same buffer. Fractions containing fimbrin were precipitated with 70% ammonium sulfate, pelleted, resuspended in 20 mM Tris, pH 7.4, and 5 mM β-Me, and applied to a 16/60 Superdex™ 200 column equilibrated in 20 mM Tris, pH 7.4, 100 mM KCl, and 5 mM β-Me. Positive fractions were pooled, diluted fourfold in 20 mM Tris, pH 7.4, and applied to a Mono Q 5/5 column equilibrated in 20 mM Tris, pH 7.4, and 5 mM β-Me. The column was eluted with a 20-ml gradient to 400 mM KCl in the same buffer. Fimbrin eluted between 180 and 200 mM KCl. α-Actinin and filamin were purified from chicken gizzard as described previously (Feramisco and Burridge, 1980).
Actin was labeled with either tetramethylrhodamine (TMR) or Alexa® 488 N-hydroxysuccinimide esters as described previously (Kellogg et al., 1988), except 0.6 M KI was used for the first depolymerization step after the labeling reaction. Arp2/3 was labeled with Alexa® 488 maleimide as described previously (Zalevsky et al., 2001). Working stocks of labeled G-actin were prepared by mixing 2 μl of a 6-mg/ml solution of G-actin labeled 1:2 to 1:3 (fluorophores/actin) with 25 μl unlabeled G-actin at 6 mg/ml.
L. monocytogenes were labeled with DAPI, killed with iodoacetamide, and stored at −80°C as described previously (Welch and Mitchison, 1998). Perfusion chambers were assembled from glass slides, glass coverslips, and double-stick tape. Bacteria were diluted into assay buffer (20 mM Pipes, pH 6.8, 75 mM KCl, 1 mM MgCl2, 1 mM ATP, and 5 mM β-Me), introduced into the chamber, and absorbed to the glass for 10 min.
For two-step reactions in total cytosol, 100,000 g brain supernatants were diluted 1:5 in assay buffer, and alexa® 488–labeled G-actin, from a working stock solution described above, was added to 5 μM. After 5 min, the solution in the chamber was replaced with the same dilution of cytosol, but with 5 μM TMR-labeled actin and 5 μM CA-GST (a gift from Terry Lechler and Rong Li, Harvard Medical School, Boston, MA).
100,000 g brain supernatants were diluted 1:5 in assay buffer and introduced into perfusion chambers containing absorbed bacteria (step 1). After a 10-min incubation at 4°C, the chamber was rinsed twice with one chamber volume each of assay buffer before introducing a solution containing 0.3 μM Arp2/3, 5 μM Alexa® 488–labeled actin, and 2 nM CapZ (step 2) in assay buffer. After 5 min, the chamber was washed twice with assay buffer followed by a solution containing 5 μM TMR-labeled actin, 5 μM CA-GST, and an appropriate dilution of fractions from bovine brain (step 3) or 1 μM recombinant fascin, fimbrin, α-actinin, or filamin. Extension of the comet tail under elongation conditions in step 3 was assessed after 5 min.
Images of the reactions were acquired on a microscope (model E800; Nikon) equipped with a cooled CCD camera (Princeton Instruments) using MetaMorph® acquisition software (Universal Imaging Corp.). Images were taken with a 60× 1.4 NA oil objective.
Purification of fascin from bovine brain
Freshly isolated bovine calf brains were homogenized in a Waring type blender in 100 ml buffer A (20 mM Pipes, pH 6.8, 25 mM KCl, 2 mM MgCl2, 5 mM β-Me, and 0.2 mM PMSF) per 100 g of tissue. Fascin was purified from a 100,000 g supernatant of bovine brain by passing the material over S-Sepharose HP (Amersham Biosciences) in buffer A. The column was eluted with a gradient from 0 to 350 mM KCl in buffer A. The flow-through contained all the activity required for step 3 and was used for purification of fascin. The bound material contained all nucleating activity and was used to purify Arp2/3. The S flow-through was dialyzed into buffer B (buffer A, but with 20 mM MES, pH 6.0) and was reapplied to the S column in buffer B and eluted with a 15-column volume gradient from 0 to 350 mM KCl. Active fractions were desalted back into buffer A on Sephadex G-25 and were applied to a heparin sulfate column equilibrated in buffer A. The column was eluted with a 15-column volume gradient from 0 to 750 ml KCl. Active fractions were dialyzed into buffer A and added to a threefold protein mass excess of phalloidin-stabilized actin filaments. The sample was incubated on ice for 60 min before collecting the actin filaments and associated proteins by centrifugation at 100,000 g for 30 min in a TLA100.3 rotor (Beckman Coulter). The pellet was washed 2× in buffer A, and was then homogenized in 1 M KCl in buffer A. The actin was pelleted and the supernatant was desalted on Sephadex G-25 into buffer A before being applied to a 1-ml Q HiTrap™ column equilibrated in buffer A. Fascin flows through the column under these conditions.
CA and AMPPNP inhibition of Arp2/3-nucleating activity
Listeria absorbed to glass in perfusion chambers were preincubated in cytosol for 10 min, and the chamber was washed twice as above. To assess inhibition of Arp2/3-dependent nucleation, the chamber was filled with 0.3 μM Arp2/3, 2 nM capping protein, and 5 μM Alexa® 488–labeled actin in assay buffer containing varying concentrations of CA-GST. Nucleation of clouds and comet tails was assessed by fluorescence microscopy after 10 min. CA inhibition of elongation was determined by adding varying amounts of CA-GST to the elongation reaction solution containing 1 μM fascin and 5 μM TMR-labeled actin. Elongation was assessed by fluorescence microscopy after 5 min. Inhibition by AMPPNP was performed in essentially the same way, except that ATP was omitted from the assay buffer during the step in which AMPPNP was used. Therefore, the only ATP in the buffer at that time comes from the actin and Arp2/3 solutions. For each concentration of CA-GST or AMPPNP, three random 40× fields were photographed in two separate experiments, and the average number of bacteria with associated clouds or tails or elongated tails was determined.
The rate of elongation in the presence and absence of fascin was determined in perfusion chambers in a three-step reaction as described above. After preincubation in cytosol, initiation in Arp2/3, actin, and capping protein, and a buffer wash, the chamber was filled with assay buffer containing 5 μM CA-GST and 5 μM actin with or without 1 μM fascin. Motility was recorded by time-lapse microscopy during this third step using either phase-contrast or differential interference contrast optics acquiring an image every 30 s. Rates and distances were determined using the “Track Objects” feature on MetaMorph®.
Motility rates as a function of G-actin concentration were determined in perfusion chambers in either a two- or three-step reaction. Bacteria were either preincubated in cytosol diluted in assay buffer as described above, or in 2 μM recombinant VASP in assay buffer with 4 mg/ml casein. After preincubation and washing, reactions were initiated with Arp2/3, capping protein, and the indicated concentration of G-actin. To determine rates in the presence of Arp2/3, motility was assessed in this second step. To determine rates in the elongation phase, the initiating solution was replaced after a 5-min incubation with fascin, CA, and the indicated concentration of G-actin.
Infection of cultured cells
For EM, PtK cells were cultured on aclar plastic. Discs of aclar were glow discharged, coated with polylysine, and washed extensively with water before the addition of PtK cells. After an overnight incubation, PtK cells were infected with L. monocytogenes for 90 min, washed into fresh media containing 50 μg/ml gentamicin, and incubated at 37°C for 4 h before fixation and processing for EM.
For immunofluorescence, BSC-1 cells or XTC cells were cultured on polylysine-coated glass coverslips, infected with L. monocytogenes for 3 h, washed into media containing 50 μg/ml gentamicin, and incubated at 37°C for an additional 4 h. The cells were fixed in methanol and costained for fascin using an anti-fascin mAb (clone 55K-2; DakoCytomation), and for actin using an anti-actin rabbit pAb (Sigma-Aldrich).
Infected PtK cells were rinsed twice with PBS and permeabilized for 3 min with 0.1% Triton X-100 in 20 mM Pipes, pH 6.8, and 100 mM KCl. The cells were then fixed with 50 mM lysine and 3% glutaraldehyde in 50 mM cacodylate, pH 7.0, for 5 min, and then with 3% glutaraldehyde in cacodylate buffer. Samples were postfixed after three rinses in cacodylate with 1% osmium and 0.8% K3Fe(CN)6 in cacodylate buffer for 15 min on ice. After three rinses in cacodylate and two rinses in water, samples were stained with 1% uranyl acetate for at least 2 h. Samples were rinsed twice with water and then were dehydrated with an ethanol series from 35 to 100% ethanol while progressively lowering the temperature from 4 to −40°C. Samples were embedded in Epon-Araldite and were thin sectioned.
For in vitro EM, motility reactions were performed in perfusion chambers consisting of a glass slide and an aclar coverslip. Bacteria were incubated in the chamber with the coverslip down to encourage absorption to the plastic coverslip. Motility reactions were performed as described above and were then fixed and processed for thin-section EM as described above.
We thank Terry Lechler, Rong Li, Paul Matsudaira, Pierre McCrea, and Fred Southwick for reagents. Mass spectrometry was performed by Jim Lee. We are grateful to members of the Mitchison lab and especially to Aaron Straight for insights and helpful discussions. Mimi Shirasu-Hiza offered needed advice in the cold room. We thank Zach Perlman and are especially grateful to Guillaume Charas for discussing mechanisms of Listeria propulsion.
W.M. Brieher acknowledges the support of the Helen Hay Whitney Foundation. This work was supported by National Institutes of Health grant GM 48027.
Abbreviations used in this paper: AMPPNP, adenosine 5′-(β,γ-imido)triphosphate; β-Me, β-mercaptoethanol; CA; cofilin homology and acidic region of N-WASP; Ptk, rat kangaroo kidney; TMR, tetramethylrhodamine.