Actin network disassembly powers dissemination of Listeria monocytogenes

ABSTRACT

Several bacterial pathogens hijack the actin assembly machinery and display intracellular motility in the cytosol of infected cells. At the cell cortex, intracellular motility leads to bacterial dissemination through formation of plasma membrane protrusions that resolve into vacuoles in adjacent cells. Here, we uncover a crucial role for actin network disassembly in dissemination of Listeria monocytogenes. We found that defects in the disassembly machinery decreased the rate of actin tail turnover but did not affect the velocity of the bacteria in the cytosol. By contrast, defects in the disassembly machinery had a dramatic impact on bacterial dissemination. Our results suggest a model of L. monocytogenes dissemination in which the disassembly machinery, through local recycling of the actin network in protrusions, fuels continuous actin assembly at the bacterial pole and concurrently exhausts cytoskeleton components from the network distal to the bacterium, which enables membrane apposition and resolution of protrusions into vacuoles.

INTRODUCTION

Several intracellular pathogens, including Listeria monocytogenes, Shigella flexneri, Rickettsia spp. and Burkholderia spp. exploit the host cell actin cytoskeleton to spread from cell to cell through the formation of membrane protrusions that resolve into double membrane vacuoles in neighboring cells (Haglund and Welch, 2011; Lambrechts et al., 2008; Stevens et al., 2006). The molecular mechanisms supporting actin-based motility have been deciphered using L. monocytogenes as a model system. Genetic investigations have led to the identification of ActA as a bacterial factor required for the polymerization of actin on the surface of cytosolic bacteria (Domann et al., 1992; Kocks et al., 1992). Biochemical studies have uncovered the central role of the ARP2/3 complex as a host cell actin nucleator in L. monocytogenes actin tail assembly (Welch et al., 1997). Combined genetic and biochemical approaches have established that ActA mimics the activity of WASP and WAVE family members, which bind the ARP2/3 complex and promote its nucleation activity (Lasa et al., 1997; Skoble et al., 2000; Welch et al., 1997; Welch et al., 1998). The actin network formed by the ARP2/3 complex is composed of short and branched filaments resulting from the binding of the complex to existing mother filaments and the nucleation of daughter filaments whose brief elongation is terminated by capping proteins (Pollard and Borisy, 2003). The expansion of the actin network formed on the surface of intracellular pathogens generates forces that propel the bacteria through the cytosol. In vitro reconstitution experiments have revealed that, in addition to the ARP2/3 complex, actin-based motility requires the presence of capping proteins and the actin-filament-severing protein cofilin, which are thought to be necessary to maintain a high-steady state level of G-actin (Loisel et al., 1999). In addition, motility is more effective in presence of profilin, α-actinin and VASP (Loisel et al., 1999). Experiments using thymus extracts have also revealed the cooperative role of cofilin, AIP1 and coronin in the disassembly of L. monocytogenes actin tails in vitro (Brieher et al., 2006).

In contrast to the mechanisms supporting cytosolic motility, the mechanisms supporting formation of membrane protrusions are poorly understood. The sequence and timing of the cellular events supporting L. monocytogenes spread from cell to cell, including the formation of membrane protrusions and their resolution into double membrane vacuoles, have been established by time-lapse microscopy (Robbins et al., 1999). After a short elongation phase, protrusions display little or no movement for variable periods of time, before they are resolved into vacuoles. The exact mechanisms supporting the elongation of the formed protrusions and their resolution into vacuoles in the neighboring cells are unknown. Structural analyses of the actin network formed in L. monocytogenes protrusions have revealed that, as opposed to the short and branched filaments observed in cytosolic tails, membrane protrusions harbor long and bundled filaments (Sechi et al., 1997). Moreover, the actin cytoskeleton factors present, such as α-actinin and ezrin, are different in cytosolic tails and in protrusions (Sechi et al., 1997). Furthermore, functional studies have revealed that there are host factors, such as the serine/threonine kinase CSNK1, that are dispensable for L. monocytogenes cytosolic motility, but required for protrusion resolution into vacuoles (Chong et al., 2011). Taken together, these observations indicate that, in spite of the fact that the minimal set of cellular factors required for reconstituting actin-based motility in vitro has been clearly defined (Loisel et al., 1999), the cellular processes supporting cell-to-cell spread through protrusion formation remain to be determined.

Here, we investigate the cytoskeleton factors required for L. monocytogenes spread from cell to cell. We uncover a crucial role for the actin network disassembly machinery in the formation and resolution of membrane protrusions.

RESULTS

Identification of host factors required for L. monocytogenes dissemination

To uncover cellular factors supporting L. monocytogenes dissemination, we developed a microscopy-based assay for quantifying the spread of GFP-expressing bacteria in a given infection focus (Fig. 1). The assay relied on the identification of spreading bacteria that tend to be scattered (Fig. 1A, Individual), as opposed to non-spreading bacteria that tend to be clustered (Fig. 1A, Clustered bacteria). We defined the spreading index as the proportion of the GFP signal corresponding to spreading bacteria in a given infection foci. We used inhibitors of the actin cytoskeleton to demonstrate that this approach allowed for the quantification of a large range of spreading defects (Fig. 1B). We next used the spreading assay to screen a siRNA library targeting genes encoding regulators of the actin cytoskeleton (Siripala and Welch, 2007) (supplementary material Table S1). The library harbored four independent siRNA duplexes targeting a given gene. We defined hits as genes for which at least two siRNA duplexes conferred spreading index values that deviated by at least 2 standard deviation units from the mean spreading index value observed in mock-treated cells. According to these selection criteria, we identified eight cytoskeleton factors required for L. monocytogenes dissemination, including components of the ARP2/3 complex, such as ARPC4, the capping protein CAPZB and the actin interacting protein 1 (AIP1, also known as WD repeat-containing protein 1, WDR1) (supplementary material Table S1, 25 nM). The specificity of the observed spreading defects was confirmed by using four independent silencing duplexes (supplementary material Table S1) and their silencing efficiency was determined at the mRNA and protein levels (supplementary material Fig. S1).

Fig. 1.

Computer-assisted image analysis of L. monocytogenes spread from cell to cell. (A) Representative examples of infection foci for mock-treated and Cytochalasin D (Cyto D) treatments (250 and 500 nM) 8 hours after infection, inhibitors were added 1 hour post-infection. (B) Titration experiments with two inhibitors of actin assembly (cytochalasin D or Latrunculin B) showing the effect of inhibitor concentration and spreading index.

Host factors required for cytosolic motility

In vitro reconstitution experiments have previously established that L. monocytogenes motility relies on the ARP2/3 complex, capping proteins and the actin-depolymerizing factor (ADF) family member cofilin-1 (CFL1) (Loisel et al., 1999). Validating our in vivo genetic screen, we identified components of the ARP2/3 complex, such as ARPC4 and the capping protein CAPZB (Fig. 2A,B). We failed to identify CFL1; however, we identified AIP1 (Fig. 2A,B, AIP1), a component of the actin network disassembly machinery that acts as a co-factor of CFL1 (Poukkula et al., 2011). To clarify the role of CFL1 and AIP1 in cytosolic motility, we determined the length of the actin tails formed in the cytosol of infected cells (supplementary material Fig. S2A). As expected for components of the disassembly machinery, we found that CFL1 or AIP1 depletion resulted in a significant increase in the length of the formed actin tails (Fig. 2C). However, the proportion of bacteria associated with F-actin (Fig. 2D) and the velocity of motile bacteria (Fig. 2E), were not affected in the cytosol of mock-treated, CFL1- or AIP1-depleted cells. We also found that co-depleting CFL1 and AIP1 resulted in a significant increase in the length of the formed actin tails (Fig. 2C), but marginally affected bacterial velocity (Fig. 2E). Finally, we localized GFP–AIP1 and CFL1–GFP to cytosolic actin comet tails (Fig. 2F). These results indicate that the CFL1 and AIP1 machinery mediates the disassembly of L. monocytogenes cytosolic tails in vivo, as suggested previously in in vitro studies (Brieher et al., 2006). However, the depletion of the components of the disassembly machinery had little impact on the bacterial velocity, probably because the actin supply in the cytosol of cells impaired for actin network disassembly was sufficient to support wild-type actin assembly at the bacterial pole.

Fig. 2.

RNAi screen for host factors involved in L. monocytogenes dissemination. (A) Images of infection foci in mock-treated (MOCK), and ARPC4-, CAPZB- and AIP1-depleted cells, after 8 hours of infection with GFP-expressing L. monocytogenes (green) and stained with DAPI (red). (B) Quantification of spreading index in cells transfected with four independent siRNA duplexes (labeled A, B, C and D) targeting ARPC4, CAPZB or AIP1. Data are presented as mean±s.e.m. of three independent experiments. The dashed line indicates the threshold corresponding to 2 s.d. units. (C,D) Length of actin tails (C), and proportion of actin-associated bacteria (D) (supplementary material Fig. S2) in the cytosol of cells in which the disassembly components, cofilin-1 (CFL1) and AIP1 have been depleted alone or in combination (Tail length: MOCK vs AIP1, P = 0.0043; MOCK vs CFL1, P = 0.0033; AIP1 versus AIP1+ CFL1, P = 0.00298; CFL1 versus AIP1+CFL1, P = 0.0356; Mann–Whitney U test). (E) Cytosolic velocity of bacteria (MOCK versus AIP1, P = 0.7431; MOCK versus CFL1, P = 0.7791; AIP1 versus AIP1+CFL1, P = 0.2131; CFL1 versus AIP1+CFL1, P = 0.4731; Mann–Whitney U test) (F) Cells were co-transfected with constructs expressing GFP–AIP1 or CFL1–GFP, and infected with CFP-expressing L. monocytogenes. AIP1 and CFL1 are enriched in the cytosolic tail. *P≤0.05; **P≤0.01; ***P≤0.001.

A role for the AIP1 and CFL1 disassembly machinery in protrusion and vacuole formation

Because the cytosolic velocity of motile bacteria was not affected in AIP1-depleted cells (Fig. 2E) and yet the bacteria failed to spread from cell to cell (Fig. 2A,B), we hypothesized that defects in the AIP1-dependent disassembly machinery might lead to defects in protrusion and/or vacuole formation. As shown in supplementary material Movie 1, L. monocytogenes forms membrane protrusions that initially elongate for a short period of time (5–10 minutes) and then display slow or no motion for variable periods of time (30–60 minutes), followed by sudden resolution into vacuoles, as previously reported (Robbins et al., 1999). We determined that AIP1 depletion led to a significant decrease in the velocity of bacteria in protrusions (Fig. 3A, AIP1). We then quantified protrusion and vacuole formation (supplementary material Fig. S2B), as previously described (Chong et al., 2011). We determined that AIP1 depletion led to an increase in the number of protrusions (Fig. 3B, AIP1 versus Mock, protrusions) that correlated with a decrease in the number of bacteria gaining access to the cytosol of adjacent cells (Fig. 3B, AIP1 versus Mock, free). To confirm the specificity of the observed phenotype, we rescued AIP1 silencing by using RNAi-resistant constructs expressing wild-type or mutant versions of AIP1 (Fig. 3C) (Mohri et al., 2006). These results showed that the rescue of the observed defects required the activity of AIP1 in the cell that formed the protrusions (Fig. 3B, AIP1 + AIP1 WT, and AIP1 + AIP1 mut). Accordingly, we found that GFP–AIP1 was enriched in protrusions when expressed in the sending cell, but was not recruited to protrusions when expressed in the receiving cell (Fig. 3D). We also established that co-depletion of AIP1 and CFL1 dramatically affected the velocity of bacteria in protrusions (Fig. 3A, AIP1 + CFL1) and the resolution of protrusion into vacuoles (Fig. 3B, AIP1 + CFL1). Similar to GFP–AIP1, CFL1–GFP was also enriched in protrusions (Fig. 3E). Taken together, these experiments uncover a crucial role for the AIP1 and CFL1 actin network disassembly machinery in protrusion and vacuole formation.

Fig. 3.

Function and localization of AIP1 and CFL1 in protrusions. (A) Velocity of elongating protrusions in mock-treated (MOCK), and ARPC4-, CAPZB- and AIP1-depleted cells (MOCK versus AIP1, P = 0.0202. MOCK versus CFL1, P = 0.2985. AIP1 versus AIP1+CFL1, P = 0.0309. CFL1 versus AIP1+CFL1, P<0.0001; Mann–Whitney U test). (B) Proportion of bacteria found in protrusions, vacuoles or free in the cytosol of neighboring cells as shown in supplementary material Fig. S3B. Cells were either mock-treated or AIP1+CFL1-depleted, in addition AIP1-depleted cells were transfected with siRNA-resistant rescue constructs expressing wild type (AIP1 +AIPWT) or depolymerization-deficient AIP1 (AIP1+AIPmut). Data are representative of three independent experiments. Proportion of protrusions: MOCK versus AIP1, P = 0.0001; MOCK versus CFL1, P = 0.2380; AIP1 versus AIP1+CFL1, P = 0.0001; AIP1 versus AIP1+AIPWT, P = 0.0001; AIP1+AIPWT versus AIP1+AIPmut, P = 0.0004. Proportion of free bacteria in secondary cell: MOCK versus AIP1, P<0.0001; MOCK versus CFL1, P = 0.2270; AIP1 versus AIP1+CFL1, P<0.0001; AIP1 versus AIP1+AIPWT, P = 0.0032; AIP1+AIPWT versus AIP1+AIPmut, P = 0.0064. All P-values are calculated using the Mann–Whitney U test. (C) Schematic representation of the AIP-1 protein. Red boxes represent WD40 domains. The position of amino acids that were mutagenized in the depolymerization-deficient mutant form of AIP1 are marked by a yellow star. (D,E) Cells were co-transfected with constructs expressing (D) membrane-targeted CFP and GFP–AIP1 or (E) membrane-targeted CFP and CFL1–GFP, and infected with CFP-expressing L. monocytogenes. AIP1 and CFL1 are enriched in the protrusion when the sending cells is transfected (white arrows). AIP1 and CFL1 are not enriched on the protrusion when the receiving cell is transfected (red arrows). Scale bars: 2 µm. *P≤0.05; **P≤0.01; ***P≤0.001.

The ARP2/3 complex is recycled in protrusions

To further understand the role of actin network disassembly in L. monocytogenes protrusions, we compared the structural and dynamic organization of the actin network in cytosolic tails and in membrane protrusions. L. monocytogenes forms a branched network relying on the incorporation of the ARP2/3 complex, resulting in a distribution of actin and ARP2/3 more intense proximal to the bacterium and slowly decreasing along cytosolic tails (Fig. 4A, Cytoplasm, and Fig. 4B).

Fig. 4.

Structural and dynamic organization of the actin network in cytosolic tails and protrusions. (A) Representative images of the actin tail and protrusion formed in cells transfected with a membrane-targeted CFP-expressing construct and infected for 4 hours with CFP-expressing L. monocytogenes. Cells were stained for F-actin (red) and ARP3 (yellow). Scale bars: 2 µm. (B,C) Distribution of ARP3–GFP or YFP–actin independently assessed in cytosolic tails (B) or in elongating protrusions (>0.01 µm/second) (C), as determined by time-lapse microscopy. Data are mean±s.e.m. (D) Dynamics of ARP3–GFP in cells transfected with membrane-targeted CFP constructs and infected for 4 hours with CFP-expressing L. monocytogenes. Gray and white arrowheads mark a reference point on the stationary (left) and elongating (right) protrusions, respectively. Scale bar: 2 µm. (E) Replica electron micrograph of a L. monocytogenes (Lm)-induced protrusion; insets display the filament organization in the proximal (blue) and distal (red) networks. Scale bars: 200 nm (main), 50 nm (insets).

The distribution of actin and ARP2/3 in elongating protrusions was similar in the network proximal to the bacterial pole (Fig. 4A, Protrusion, and Fig. 4C). However, analyses of the network distal to the bacterial pole revealed a dramatic depletion of ARP2/3 with respect to actin (Fig. 4A, Protrusion, and Fig. 4C). Time-lapse microscopy revealed that the ARP2/3-containing network formed at the bacterial pole was subsequently remodeled into an ARP2/3-devoid distal network during protrusion elongation (Fig. 4D, and supplementary material Movie 2). In agreement with this notion, electron microscopy images of protrusions demonstrated the presence of a branched network at the bacterial pole along with variable amounts of long and bundled filaments (Fig. 4E, proximal). This was in contrast with the distal network, which was essentially composed of long and bundled filaments (Fig. 4E, distal), as previously reported (Sechi et al., 1997). Fluorescence recovery after photo bleaching experiments (FRAP) revealed a minimal recovery of the pool of ARP2/3 located at the bacterial pole in protrusions, demonstrating very little contribution from the cytosol (supplementary material Fig. S3A). Collectively, these results demonstrate that protrusion elongation relies on a pool of ARP2/3 that is constantly recycled from the distal network and incorporated into the newly formed network at the bacterial pole.

Actin network recycling in elongating protrusions

We further examined the dynamics of the actin network by using a photo-activatable version of GFP fused to mCherry–actin (Welman et al., 2010). In the cytosol, photo-activation of the distal region of an elongating actin tails produced a GFP signal that slowly decreased in intensity as the tail disassembled (Fig. 5A, white arrowhead in tiled images, red dashed line in kymograph; supplementary material Movie 3) (Theriot et al., 1992). In elongating protrusions, and similar to the situation observed in the cytosol, photo-activation of the distal network produced a GFP signal that slowly decreased in intensity at the site of photo-activation (Fig. 5B, white arrowhead in tiled images, red dashed line in kymograph; supplementary material Movie 4). In contrast with the situation observed in the cytosol (Fig. 5A), we observed the appearance of a signal at the bacterial pole that increased in intensity as protrusions elongated (Fig. 5B, blue arrowhead in tiled images, black dashed line in kymograph; supplementary material Movie 4). Measurement of the diffusion rate of photo-activatable free GFP and GFP–actin in protrusions (supplementary material Fig. S3B–D) and treatment with actin polymerization inhibitors (supplementary material Fig. S3E) revealed that (1) the disappearance of photo-activated molecules at the site of photo-activation reflects actin network disassembly, and (2) the accumulation of photo-activated molecules at the bacterial pole, reflects actin assembly. In addition, FRAP experiments with GFP and GFP–actin indicated very slow recovery of the pool of actin in protrusions, confirming the notion that protrusions constitute a confined environment (supplementary material Fig. S3A). Thus, similar to the situation observed with ARP2/3, these experiments indicate that actin molecules are recycled from the distal network in the confined environment of protrusions and fuel assembly of the network proximal to the bacterial pole in elongating protrusions.

Fig. 5.

Disassembly of the distal network in L. monocytogenes protrusions fuels actin assembly at the bacterial pole. (A–C) Time-lapse imaging of photo-activation of β-actin fused to mCherry and photo-activatable GFP in a cytosolic tail (A), an elongating protrusion (B) and a stationary protrusion (C). White arrows indicate the site of photo-activation and blue arrows indicate the initial position of the bacterial pole. Scale bar: 2 µm. Kymographs represent the evolution of the photo-activated signal over space (x-axis) and time (y-axis). Red dashed lines indicate the progression of the signal from the initial site of photo-activation and black dashed lines indicate the signal evolution at the bacterial pole. (D) Percentage of the initial photo-activated signal trafficked to the bacterial pole after 35 seconds in elongating and stationary protrusions (<0.01 µm/s) (stationary versus elongating, P = 0.6097, Mann–Whitney U test). (E) Negative correlation of elongation rate and retrograde flow in protrusions (R = −0.7375, P<0.0001, Spearman's rank correlation).

Actin network recycling in stationary protrusions

In addition to elongating protrusions, we also examined the dynamics of the actin network in protrusions that became stationary after their elongating phase and before their resolution into vacuoles. Unexpectedly, we found that, in those protrusions that displayed little or no movement (Fig. 5C, blue arrowhead in tiled images, black dashed line in kymograph; supplementary material Movie 5), the rate of actin assembly was similar to the rate observed in elongating protrusions (Fig. 5D). In addition, as the corresponding signal decreased in intensity, photo-activated molecules drifted away, as a whole, from the site of photo-activation (Fig. 5C, white arrow in tiled images, red dashed line in kymograph; supplementary material Movie 5). These experiments revealed that, as a result of network assembly at the bacterial pole, the whole network was displaced backwards in stationary protrusions. In agreement with this notion, treatment with actin polymerization inhibitors dramatically affected actin assembly at the bacterial pole (supplementary material Fig. S3E) and the observed backward drift of the actin network (supplementary material Fig. S3F). We refer to this phenomenon as ‘retrograde flow in protrusions’, based on analogy to the actin retrograde flow observed in lamellipodia (Forscher and Smith, 1988; Pollard and Borisy, 2003; Wang, 1985). We further demonstrated an inverse correlation between elongation and retrograde flow in protrusions (Fig. 5E). Thus, actin assembly at the bacterial pole mediates (1) protrusion elongation with minimal retrograde flow during the initial period of protrusion formation, and (2) maximal retrograde flow in protrusions that no longer elongate.

AIP1- and CFL1-dependent recycling of the actin network in protrusions

We next investigated the role of the AIP1- and CFL1-dependent disassembly machinery in the dynamics of the actin network in protrusions. We determined that the distal region of the protrusions formed in AIP1- and AIP1- plus CFL1-depleted cells was wider (Fig. 6A,B), and displayed altered distribution of F-actin (Fig. 6A,C) and ARP2/3 (Fig. 6A,D), suggesting a role for the AIP1 and CFL1 disassembly machinery in the recycling of the distal network. We used the photo-activation assay to quantify the disappearance of photo-activated molecules at the site of photo-activation, as a measure of actin network disassembly, and accumulation of photo-activated molecules at the bacterial pole, as a measure of actin assembly. We found that the protrusions formed in AIP1- and AIP1- plus CFL1-depleted cells displayed slow actin network disassembly (Fig. 6E), which correlated with slow actin assembly at the bacterial pole (Fig. 6F). Thus, defects in the AIP1- and CFL1-dependent-machinery led to severe defects in the recycling of the actin network formed by the bacteria in protrusions, which dramatically affected actin assembly at the bacterial pole.

Fig. 6.

Role of AIP1, CFL1 and GMFB in L. monocytogenes protrusions. (A) Representative images of protrusions formed in mock-treated (MOCK), and AIP1-, AIP1+CFL1- or AIP1+GMFB-depleted cells transfected with a membrane-targeted CFP-expressing construct and infected for 4 hours with CFP-expressing L. monocytogenes. Scale bars: 2 µm. (B–D) Width of protrusions (B), distribution of F-actin (C) and ARP2/3 (D) in protrusions as shown in A. Data are mean±s.e.m. (E) Percentage of signal disappearance of initial photo-activated signal after 35 seconds in mock-treated, AIP1-, AIP1+CFL1- or AIP1+GMFB-depleted cells (MOCK versus AIP1, P = 0.0016; AIP1 versus AIP1+CFL1, P = 0.0077; AIP1 versus AIP1+GMFB, P = 0.0365; Mann–Whitney U test). (F) Percentage of the initial photo-activated signal trafficked to the bacterial pole after 35 seconds in Mock-treated, AIP1-, AIP1+CFL1- or AIP1+GMFB-depleted cells (AIP1 versus MOCK, P<0.0001; AIP1 versus AIP1+CFL1, P<0.0001; AIP1 versus AIP1+GMFB, P = 0.0005; Mann–Whitney U test). (G) Proportion of bacteria found in protrusions, vacuoles or free in the cytosol of neighboring cells. Proportion of protrusions: AIP1 versus AIP1+GMFB, P = 0.0016. Proportion of free bacteria in secondary cell: AIP1 versus AIP1+GMFB, P<0.0001. All P-values are calculated using the Mann–Whitney U test. *P≤0.05; **P≤0.01; ***P≤0.001.

Identification of GMFB as a component of the AIP1-dependent disassembly machinery

To define additional components of the AIP1-dependent disassembly machinery, we screened the cytoskeleton library for factors whose depletion would enhance the spreading defect phenotype displayed by AIP1-depleted cells. This approach confirmed the involvement of CFL1 and led to the identification of glial maturation factor β (GMFB), twinfilin 2 (TWF2) and adenylate cyclase-associated protein 1 (CAP1) (supplementary material Fig. S4 and Table S1, 25 nM+25 nM AIP1). The specificity of these genetic interactions (supplementary material Table S1) was confirmed by using various combinations of independent and validated silencing reagents (supplementary material Figs S1, S4). We further investigated the role of GMFB, which together with CFL1 and TWF2, is a member of the ADF family (Nakano et al., 2010). We found that, reminiscent of the situation observed in AIP1- plus CFL1-depleted cells, the distal region of the protrusions formed in AIP1- plus GMFB-depleted cells was wider (Fig. 6B) and displayed altered distribution of F-actin (Fig. 6A,C) and ARP2/3 (Fig. 6A,D). We also established that the protrusions formed in AIP1- plus GMFB-depleted cells displayed slow actin network disassembly (Fig. 6E), which correlated with slow actin assembly at the bacterial pole (Fig. 6F). We finally counted the membrane protrusions and double-membrane vacuoles in neighboring cells and determined that AIP1 plus GMFB depletion led to a dramatic accumulation of protrusions that correlated with a decrease in the number of bacteria gaining access to the cytosol of adjacent cells (Fig. 6G, free). Collectively, these results define GMFB as a component of the AIP1-dependent disassembly machinery required for local recycling of the actin network in protrusions, which is crucial for bacterial dissemination.

DISCUSSION

Seminal studies on L. monocytogenes have uncovered an essential role for the actin assembly machinery in actin-based motility. However, the cellular processes supporting bacterial spread from cell to cell have remained elusive. Here, we have uncovered an essential role for the disassembly machinery in the formation and resolution of membrane protrusions in vivo. On the basis of our findings, we propose the following model for how L. monocytogenes spreads from cell to cell. The bacterium first develops actin-based motility in the cytosol of infected cells and initiates protrusion formation as it encounters the plasma membrane. The AIP1-dependent disassembly machinery (Fig. 7A) then powers the elongation process by recycling the network formed by the bacterium, thereby fueling ARP2/3-dependent actin assembly at the bacterial pole (Fig. 7B). As the recycling process exhausts the cytoskeleton components from the distal network, protrusions become stationary, and actin assembly results in retrograde flow. Exhaustion of the cytoskeleton from the distal network together with the simultaneous generation of forces by actin assembly at the bacterial pole, and the resulting retrograde flow, allow for concerted membrane apposition and membrane tensions, that lead to membrane disruption and vacuole formation (Fig. 7C).

Fig. 7.

Model of local recycling of the actin network in L. monocytogenes protrusions. (A) Components of the AIP1-dependent disassembly machinery (CFL1, GMFB, TWF2 and CAP1) whose depletion enhances the spreading defect phenotype displayed by AIP1-depleted cells. (B) The bacterial factor ActA (green dots) promotes the nucleation activity of the ARP2/3 complex (red dots), which leads to the assembly of a branched network at the bacterial pole (blue lines and red dots). As protrusions elongate, the AIP1-dependent disassembly machinery recycles G-actin and ARP2/3 from the distal network, which fuels continuous F-actin assembly at the bacterial pole (red arrow). (C) The life cycle of protrusions can be divided in four phases: (1) Emerging protrusions, actin assembly propels the cytosolic bacterium against the plasma membrane, which protrudes into the adjacent cell; (2) elongating protrusions, as protrusions elongate, the disassembly machinery recycles the distal network thereby fuelling further assembly at the bacterial pole in this confined system; (3) stationary protrusions, as the recycling process exhausts the cytoskeleton components from the distal region of protrusions, protrusions become stationary, and continuous actin assembly results in retrograde flow; and (4) protrusion-to-vacuole transition, complete exhaustion of the cytoskeleton components from the distal network allows for membrane apposition in the distal region of protrusions. Continuous generation of forces due to actin assembly and retrograde flow leads to membrane disruption and resolution of the protrusion into a double-membrane vacuole.

In addition to the mechanisms supporting bacterial dissemination, our work further establishes striking parallels between the cellular processes supporting the formation of L. monocytogenes protrusions and cellular structures, such as lamellipodia (Haglund and Welch, 2011; Lambrechts et al., 2008; Stevens et al., 2006). This includes the conserved role of essential components, such as AIP1 and CFL1 (Poukkula et al., 2011; Siripala and Welch, 2007), and the potent actin network retrograde flow resulting from actin assembly (Forscher and Smith, 1988). Importantly, our genetic investigations led to the identification of cytoskeleton factors, such as GMFB, TWF2 and CAP1 that genetically interact with AIP1. Similar to CFL1, GMFB and TWF2 are members of the ADF family. Mammalian TWF1, and potentially TWF2, displays capping and severing activity in vitro (Poukkula et al., 2011). Yeast (Saccharomyces cerevisiae) Gmf1, a homolog of GMFB, displays ARP2/3-specific de-branching activity (Gandhi et al., 2010; Nakano et al., 2010), and its in vitro debranching activity was recently shown to be conserved in mammalian GMFγ (Ydenberg et al., 2013). CAP1 is thought to participate in the dissociation of the cofilin–ADP-actin complex (Mattila et al., 2004) and was recently shown to directly enhance cofilin-mediated filament severing (Chaudhry et al., 2013; Normoyle and Brieher, 2012). L. monocytogenes protrusions therefore are an attractive model system to decipher how the activities of AIP1, CFL1, GMFB, TWF2 and CAP1 contribute to the dynamics of the actin cytoskeleton in vivo.

MATERIALS AND METHODS

Bacterial and mammalian cell growth conditions

Listeria monocytogenes strain 10403S was grown overnight in brain heart infusion (BHI) (Difco) supplemented with 10 µg/ml erythromycin (Gibco) at 30°C without agitation prior to infection. HeLa 229 cells (ATCC #CCL-2.1) cells were grown in high-glucose Dulbecco's modified Eagle's medium (DMEM) (Invitrogen) supplemented with 10% fetal bovine serum (FBS) (Gibco) at 37°C in a 5% CO₂ incubator.

L. monocytogenes infection

Cells were infected with L. monocytogenes strain 10403S expressing GFP or CFP under the control of an IPTG-inducible promoter. The plates were centrifuged at 200 g for 5 minutes and internalization of the bacteria was allowed to proceed for 1 hour at 37°C before gentamicin (50 µM final) and IPTG (10 mM final) were added in order to kill the remaining extracellular bacteria and induce expression of fluorescent protein genes in internalized bacteria.

Mammalian cell transfection

HeLa 229 cells were transfected using the X-treme gene HP reagent (Roche Applied Science) 24 hours prior to infection. A list of DNA constructs is given in supplementary material Table S2.

Immunofluorescence staining procedures

Cells were seeded on glass coverslips and infected with L. monocytogenes. At the appropriate time points, cells were fixed for 20 minutes with 4% formaldehyde in PBS, permeabilized with 0.1% Triton X-100 for 5 minutes, blocked for 45 minutes in 3% BSA and incubated overnight at 4°C with primary antibody. Cells were washed in PBS and incubated for 1 hour with the secondary antibody. The samples were then mounted onto glass slides using DABCO antifade reagent. A list of antibodies and concentrations used is given in supplementary material Table S2.

Epifluorescence microscopy

Epi-fluorescence images were acquired using a TE 2000 microscope (Nikon) equipped with a 60× objective (Nikon). Image acquisition and analysis was conducted using the MetaMorph 7.1 software (Molecular Devices).

Confocal microscopy

Images of infected cells were acquired with a Nikon TE2000 spinning disc confocal equipped with a live-cell apparatus and a Micropoint bleaching laser microscope at 37°C in 5% CO₂ (Andor Technology). Analyses were performed with the Volocity software package (Improvision).

Replica electron microscopy

Platinum replica electron microscopy was performed, as described previously (Svitkina, 2007), using an extraction buffer containing 1% Triton X-100, 2 µM phalloidin, 100 mM PIPES pH 6.9, 1 mM EGTA and 1 mM MgCl₂. Before mounting on grids, replicas were additionally treated with Clorox bleach, as described previously (Svitkina and Borisy, 2006), to digest organic material and improve clarity of samples. Samples were analyzed using JEM 1011 transmission electron microscope (JEOL USA, Peabody, MA) operated at 100 kV. Images were captured with an ORIUS 832.10W CCD camera (Gatan, Warrendale, PA) and presented in inverted contrast.

High-throughput imaging and computer-assisted image analysis

384-well plates were imaged using a TE 2000 microscope (Nikon) equipped with a Orca ER Digital CCD Camera (Hamamatsu), motorized stage (Prior), motorized filter wheels (Sutter Instrument, Inc.) and a 10× objective (Nikon) mounted on a Piezo focus drive system (Physik Instrumente). Image acquisition and analysis were conducted using the MetaMorph 7.1 software (Molecular Devices, Inc.). To identify foci of infection, images corresponding to the GFP channel were first thresholded in order to identify objects corresponding to bacteria (see Fig. 1). Infection foci were delineated using the ‘close’ morphology filter of the MetaMorph software (Identification of infection focus, green objects). This step defined regions of interest (ROI) corresponding to infection foci (Fig. 1, thresholded image, yellow line). Within a given ROI, we next used the image morphometry analysis (IMA) module of the MetaMorph software to quantify (1) [Tot] as the total gray intensity values corresponding to the GFP signal, and (2) [Ind] as the gray intensity values corresponding to individual bacteria (Individual bacteria bottom panels). The spreading index, [Ind]/[Tot], represents the proportion of GFP signal represented by individual (and therefore spreading) bacteria within a given infection focus.

RNAi and validation procedures

Cells were transfected by reverse transfection with Dharmafect1 and individual siRNA duplexes (A, B, C and D, 50 nM final) (Dharmacon) or a pool of the four silencing reagents (12.5 nM each, 50 nM total) and incubated for 72 hours. The references of individual siRNA products are given in supplementary material Table S1. For real-time PCR analysis, total RNA and first-strand cDNA synthesis was performed using the TaqMan gene expression Cells-to-Ct kit (Applied Biosystems) as recommended by the manufacturer. Primers were designed using Roche Universal ProbeLibrary Assay Design Center. mRNA was quantified for individual genes as well as a GAPDH internal control using the LightCycler 480 Master Kit and a LightCycler 480 instrument (Roche Biochemicals, Indianapolis, IN). Primers and probe number are given in supplementary material Table S2. For western blot analysis, cells were transfected with the silencing reagents in a 24-well or a 6-well plate format and lysed after 3 days directly in Laemmli sample buffer. A list of antibodies and concentration used is shown in supplementary material Table S2.

Cytosolic actin tail length and F-actin association

Mock- and siRNA-treated HeLa 229 cells were infected for 4 hours, fixed, permeabilized and stained with Alexa-Fluor-568-conjugated phalloidin (1∶500) (Invitrogen) and Hoechst 33342 (1∶500) (Invitrogen) for 1 hour. Epifluorescence images were acquired, the proportion of bacteria associated with F-actin in the form of a cloud or a polarized tail (supplementary material Fig. S2A) and the length of polarized tails were measured (supplementary material Fig. S2A).

Cytosolic motility

Mock- and siRNA-treated HeLa 229 cells were seeded on day 0 on 35-mm imaging dishes (MatTek). Cells were transfected on day 2 with a given construct (supplementary material Table S2). Cells were infected on day 3 and imaged exclusively from 3 to 6 hours post-infection. Primary infected cells transfected with pDsRed-Monomer-Mem (Clontech) were imaged for 10 minutes with a 444 nm laser (CFP bacteria) and a 561 nm laser (Ds-Red membrane). z-stacks spanning the entire cell were acquired every 30 second with a spacing of 0.5 µm in the z dimension. The Volocity software package was used to identify bacteria as objects and track them in four dimensions. Each track was manually validated. A speed was obtained for no less than 15 bacteria per treatment, originating from ten cells in three independent experiments.

Protrusion formation and resolution into vacuoles

Mock- and siRNA-treated HeLa 229 cells were seeded on day 0. Cells were transfected with pDsRed-Monomer-Membrane DNA construct (Clontech) on day 2 cells were infected for 5 hours before fixation. Epifluorescence images were acquired and the proportion of bacteria found in protrusions, vacuoles or free in neighboring cells was assessed for no less than 15 primary infected cells in three independent replicates.

Protein localization in protrusions

HeLa 229 cells were seeded on day 0. Cells were transfected with a given construct (supplementary material Table S2) on day 2. Cells were infected on day 3 and fixed after 5 hours. The cells were counterstained with Alexa-Fluor-568-conjugated phalloidin and 5 individual protrusions were imaged per construct. Analyses were restricted to protrusions above 7 µm in length and which had a horizontal orientation (distributed in less than five 0.5 µm z-stack slices). A line profile was calculated from a 0.5-µm-wide line manually drawn on each protrusion from the bacterial pole to the distal end of the protrusions. Line profile intensities were calculated for three channels, corrected to the background fluorescence for each channel and normalized to maximum intensity.

Tracking of proteins during protrusion elongation

Primary infected cells, transfected with a membrane–CFP plasmid (supplementary material Table S2) and a GFP- or YFP-tagged protein construct (supplementary material Table S2), were imaged for 10 minutes with the 444 nm laser and the 490 nm laser. CFP-expressing bacteria and the membrane CFP were acquired simultaneously to avoid excessive photo-bleaching. z-stacks spanning the entire cell were acquired every 40 seconds with a spacing of 0.5 µm in the z dimension. The Volocity software package (Perkin Elmer) was used to correct for photo-bleaching. Individual bacteria undergoing continuous movement (>0.01 µm/second) for at least five successive time points (200 seconds) were identified and selected for analysis. The analysis was restricted to protrusions above 7 µm in length and which had a horizontal orientation (distributed in less than five 0.5 µm z-stack slices). A line profile was calculated from a 0.5-µm-wide line manually drawn on each protrusion and each time point from the bacterial pole to the distal end of the protrusion. Line profile intensities were calculated and corrected with the background fluorescence for each channel. Relative intensity figures were computed by calculating the relative intensity of a signal along the line profile as compared to the maximum intensity in the protrusion at a given time point and for a given fluorophore. Protein distribution profiles were generated for no less than six protrusions in three independent experiments.

Photo-activation and photo-bleaching

Primary infected cells were transfected with a membrane-CFP plasmid (supplementary material Table S2) and a PAGFP–mCherry-tagged β-actin construct (Welman et al., 2010). A Micropoint 404 nm laser (Andor Technology) was used to photo-activate a region of interest (ROI) situated ∼2 µm from a bacterial pole in protrusions (above 7 µm in length). The ROI surface was kept constant in all experiments. Two pre-activation images were acquired, followed by two recovery images at 2.5-second intervals. The cells were then imaged every 5 seconds for a total of 35 seconds. Laser power and exposure times were kept to a minimum (<50%, <200 milliseconds) to avoid non-specific photo-activation. The localization (cytoplasmic, protrusion) and length of protrusions were assessed visually after the photo-activation experiment by using the membrane CFP marker. The mCherry signal was used to determine the rate of protrusion elongation at the bacterial pole. Images were analyzed using the ImageJ software. The trajectory was drawn along the mCherry signal and used to measure the relative signal intensity profiles of photo-activated molecules corresponding to the first and last image post-activation (35 s). A kymograph was generated using all images of a given trajectory and used to calculate the speed of retrograde flow.

Photo-bleaching experiments were carried out in cells expressing various fluorescent markers (supplementary material Table S2). Two pre-bleaching images were acquired before the area immediately adjacent to bacteria in protrusion was photo-bleached, and recovery images were captured at maximum speed for 10 seconds. Images in which the bacteria moved outside the constant ROI during the experiment were disregarded. FRAP analysis was conducted with the Volocity software package using a single constrained exponential setting. Three separate independent recovery profiles were generated for each fluorescent marker.