The formin FHOD1 and the small GTPase Rac1 promote vaccinia virus actin–based motility

The life cycle of Orthopoxviruses, such as vaccinia virus, relies on the dissemination of two infectious forms, the intracellular mature viruses (IMVs) and the extracellular viruses (EVs). IMVs propagate after massive release by cell lysis at the end of the infectious cycle. Vaccinia dissemination is also supported by the spread of EVs from cell to cell through actin-based motility (Smith et al., 2002; Roberts and Smith, 2008). The formation of EVs is complex and first requires wrapping of IMVs in double membranes derived from the trans-Golgi network or endosomes, leading to the production of intracellular enveloped viruses (IEVs; Hiller and Weber, 1985; Tooze et al., 1993; Schmelz et al., 1994). IEVs travel on microtubules and the fusion of their outer membrane to the plasma membrane leads to formation of cell-associated EVs (CEVs; Hollinshead et al., 2001; Rietdorf et al., 2001; Ward and Moss, 2001a; van Eijl et al., 2002). The viral proteins B5 and A36 are essential for vaccinia actin–based motility (Sanderson et al., 1998; Wolffe et al., 1998; Katz et al., 2003). B5 and A36 are located in the inner and outer membrane of IEVs, respectively (Engelstad et al., 1992; Isaacs et al., 1992; Parkinson and Smith, 1994; Röttger et al., 1999). The fusion of the outer membrane of IEVs to the plasma membrane exposes B5 at the surface of CEVs (Engelstad et al., 1992; Isaacs et al., 1992) and positions A36 in the plasma membrane, underneath CEVs (van Eijl et al., 2000). The exact role of B5 in actin-based motility is unknown. The phosphorylation of the cytosolic tail of A36 by Src and Abl family members creates docking sites for the adaptor proteins Nck1 and Grb2 (Scaplehorn et al., 2002; Newsome et al., 2006). Nck1 and Grb2 mediate the recruitment of N-WASP, which mediates the recruitment and activation of the ARP2/3 complex (Frischknecht et al., 1999). The ARP2/3 complex binds existing mother actin filaments and nucleates the formation of daughter filaments that branch out from the mother filament leading to the formation of a branched network (Pollard and Borisy, 2003). The origin of the mother filaments required for initiation of ARP2/3-mediated actin polymerization is unknown. The assembly of actin beneath CEVs generates forces that drive the formation of virus-tipped membrane protrusions also referred to as actin tails, which leads to viral release into the extracellular environment or viral transmission to neighboring cells (Smith et al., 2002; Roberts and Smith, 2008).

The manipulation of the actin cytoskeleton represents a common theme in pathogen dissemination. Similar to vaccinia virus, several bacterial pathogens, such as Listeria monocytogenes and Shigella flexneri also manipulate the ARP2/3 complex in order to spread from cell to cell. L. monocytogenes recruits and activates the ARP2/3 complex through ActA (Welch et al., 1998), a bacterial factor that mimics the activity of the nucleation-promoting factor N-WASP (Lasa et al., 1997; Skoble et al., 2000; Boujemaa-Paterski et al., 2001; Chong et al., 2009). In contrast, S. flexneri expresses a bacterial adaptor, IcsA, that mediates the recruitment of N-WASP, which then recruits and activates the ARP2/3 complex (Bernardini et al., 1989; Goldberg et al., 1994; Suzuki et al., 2002). On the basis of these investigations, pathogen dissemination was thought to rely solely on the manipulation of the ARP2/3 complex. However, Heindl et al. (2010) have recently reported that, in addition to manipulating the ARP2/3 complex for displaying cytosolic motility, S. flexneri requires formin family members for plasma membrane protrusion formation. Moreover, recent studies have unexpectedly revealed that Rickettsia spp. do not require the ARP2/3 complex to form actin tails and instead express and use the bacterial factor Sca2, a functional mimic of mammalian formins (Haglund et al., 2010; Kleba et al., 2010). Furthermore, a role for N-WASP recycling and clathrin-mediated endocytosis was recently reported in vaccinia actin–based motility (Weisswange et al., 2009; Humphries et al., 2012). Thus, the cellular requirements and related functions supporting pathogen dissemination may prove to be more complex than previously depicted.

Here, we uncover the formin FHOD1 and the small GTPase Rac1 as novel regulators of vaccinia actin tail formation. Our work suggests that vaccinia virus has evolved a complex strategy of actin-based motility relying on the integration of the activity of the ARP2/3 complex and the formin FHOD1.

We identified the formin FHOD1 as a cellular factor required for vaccinia dissemination by conducting the following genetic approach. We first developed an assay for quantification of virus spread in human epithelial cells using automated fluorescence microscopy and computer-assisted image analysis. Wild-type virus expressing a B5R-GFP chimera spread from primary infected cells to neighboring cells at 16 h after infection (Fig. 1, A and B, vB5R-GFP and control). This time point was selected on the basis of preliminary time course experiments indicating that at16 h after infection, vaccinia dissemination mainly relied on ARP2/3-dependent actin-based motility (Fig. S1 Ahttp://www.jcb.org/cgi/content/full/jcb.201303055/DC1). As expected, the recombinant ΔA36R vaccinia variant deficient in actin tail formation led to the formation of infection foci of small size, reflecting a severe spreading defect (Fig. 1, A and B, ΔA36R). We used computer-assisted image analysis to quantify the size of the formed foci (Fig. 1 A, bottom). The vast majority of the foci formed in cells infected with the ΔA36R vaccinia variant displayed a size corresponding to single infected cells, which we defined as small (Fig. 1, ΔA36R, green). We also observed foci significantly larger in size that presumably resulted from the merge of two or more small foci (Fig. 1, ΔA36R, white). In cells infected with wild-type viruses, we observed a mixed population of foci whose size ranged from small, as defined by the size of the foci formed in cells infected with the ΔA36R vaccinia variant, to large (Fig. 1, control, green and white). To implement automated image analysis procedures, we computed the spreading index as the ratio [total foci − small foci]/[total foci], which accurately reflected the observed enrichment in foci of large size within the population of foci formed in cells infected with wild-type virus (Fig. S1 B). Statistical analyses showed that the mean spreading index was 0.36 ± 0.08 for the wild-type virus and 0.13 ± 0.03 for the ΔA36R variant, reflecting highly significant differences in the spreading ability of the two viral strains (Fig. 1 B). We next determined whether our assay would allow us to identify host factors involved in cell-to-cell spread by silencing the expression of N-WASP and ARPC4. ARPC4 is a component of the ARP2/3 complex and N-WASP is the nucleation-promoting factor that recruits and activates the ARP2/3 complex beneath CEVs. As expected, and similar to the deletion of A36R, depletion of N-WASP or ARPC4 resulted in impaired actin-based motility, as reflected by spreading indexes of 0.13 ± 0.03 and 0.17 ± 0.05, respectively (Fig. 1, A and C).

We then assembled a siRNA library covering regulators of the actin cytoskeleton (Siripala and Welch, 2007). The screening of this library using the assay described in Fig. 1 led to the identification of FHOD1 as a cytoskeleton factor required for vaccinia virus dissemination. We validated the specific involvement of FHOD1 by using four independent silencing reagents targeting FHOD1 expression (Fig. 2, A and B, FHOD1siA–D). We found that siRNA duplexes A, B, and D conferred the best silencing efficiency at the transcriptional level (Fig. 2 C) as well as the translational level (Fig. 2 D). These siRNA duplexes also conferred the best spreading defect phenotypes (Fig. 2, A and B). We further showed that treatment with FHOD1 siRNA duplexes did not confer general cell growth defect (Fig. S2 A). Finally, we established that FHOD1-depleted cells could support the dissemination of unrelated pathogens, such as L. monocytogenes (Fig. S2 B). Altogether, these results indicate that the formin FHOD1 is specifically required for vaccinia virus dissemination.

We next determined whether the dissemination defect observed in FHOD1-depleted cells originated from a defect in actin-based motility. To this end, we infected cells with a recombinant vaccinia virus variant expressing an mCherry-tagged version of the viral core protein A5L and a triple FLAG–tagged version of B5R (vA5L-mCherry/3xF-B5R; Alvarez and Agaisse, 2012). FLAG immunostaining without permeabilization was used to detect and quantify the number of CEVs displaying actin tails. We found that depletion of FHOD1 resulted in a 2.5-fold decrease in the number of CEVs displaying actin tails (Fig. 3, A–C). We also used time-lapse microscopy to monitor vaccinia virus actin–based motility in cells expressing YFP-N-WASP and CFP-actin. The velocity of N-WASP–tipped actin tails was determined in mock-treated or FHOD1-depleted cells. We found that silencing FHOD1 resulted in a 30% decrease in the speed of the virus compared with mock-treated cells (Fig. 3 D). Collectively, these observations indicate that FHOD1 contributes to the initiation and the elongation of vaccinia virus actin tails.

We next determined whether the decrease in actin tail formation observed in FHOD1-depleted cells reflected a defect in the recruitment of N-WASP beneath CEVs. To this end, we analyzed endogenous N-WASP localization in cells infected with the recombinant vA5L-mCherry/3xF-B5R vaccinia variant (Fig. 4 A). We found that, in FHOD1-depleted cells, N-WASP was recruited to CEVs as efficiently as in mock-treated cells (Fig. 4 B). Because we previously observed a 2.5-fold decrease in actin tail formation in FHOD1-depleted cells (Fig. 3 C), these results suggested that, as opposed to the situation observed in mock-treated cells, a significant proportion of CEVs successfully recruited N-WASP and yet did not display actin tails in FHOD1-depleted cells. To further confirm this possibility, we quantified the number of N-WASP–positive (A36-positive) CEVs displaying actin tails in mock-treated and FHOD1-depleted cells (Fig. 4 C). We found that >75% and <25% of N-WASP–positive CEVs were associated with actin tails in mock-treated and FHOD1-depleted cells, respectively (Fig. 4 D). These results indicate that FHOD1 is not required for the recruitment of N-WASP to CEVs, but contributes to the efficiency of N-WASP–dependent actin tail formation.

We next analyzed the localization of FHOD1 in virus-induced actin tails. Immunostaining of endogenous FHOD1 revealed colocalization with virus-tipped actin tails in infected cells (Fig. S3 A). To evaluate the specificity of FHOD1 recruitment to actin tails, we analyzed the localization of FHOD1 and mDia1 in vaccinia-infected cells. Both FHOD1 and mDia1 belong to the group of diaphanous related formins and share a similar functional domain organization. Cells expressing GFP-FHOD1 or GFP-mDia1 were infected with vaccinia virus and actin tails were scored positive or negative for the corresponding GFP fusion proteins. We observed that >90% of vaccinia virus actin tails were positive for FHOD1, whereas <25% of the tails were positive for mDia1 (Fig. S3 B). These results indicate that FHOD1 specifically localizes to vaccinia actin tails.

We next investigated the determinants supporting FHOD1 actin tail localization. The formin FHOD1 is organized into an N-terminal GTPase binding domain (GBD), the diaphanous autoregulatory domain (DAD) recognition domain FH3, a short proline-rich FH1 domain, an FH2 domain, and the C-terminal DAD (Schulte et al., 2008). The FH2 domain confers to formins the ability to nucleate, elongate, and cap actin filaments at barbed ends (Watanabe et al., 1999; Pruyne et al., 2002; Li and Higgs, 2003; Harris et al., 2004; Moseley et al., 2004, 2006; Otomo et al., 2005). These barbed end activities rely on a conserved isoleucine residue within the FH2 domain (isoleucine 705 in FHOD1; Harris et al., 2006). The FH1 domain is required for binding profilin-actin and may also interact with SH3 or WW domain–containing proteins (Pring et al., 2003; Romero et al., 2004). Interaction between FH3 and DAD domains folds formins into an autoinhibited conformation (Alberts, 2001; Lammers et al., 2005; Nezami et al., 2006; Schönichen et al., 2006). Binding of small GTPases of the Rho/Rac/Cdc42 family to the GBD domain relieves the autoinhibited conformation and leads to formin activation (Wallar and Alberts, 2003). It has been previously shown that FHOD1 activity requires intact GBD, FH1, and FH2 domains (Takeya and Sumimoto, 2003; Schulte et al., 2008). To further characterize the role of the functional domains of FHOD1 in the formation of vaccinia virus actin tail formation, we first generated a full-length siRNA-resistant form of FHOD1 fused to dsRed. We showed that expression of this siRNA-resistant form of FHOD1 (Fig. 5, A and B, FHOD1si, wt) rescued the actin tail formation defect observed in FHOD1-depleted cells (Fig. 5, A and B, FHOD1si, –). We next determined whether siRNA-resistant forms of FHOD1, bearing deletions of the GBD, FH1, or FH2 domains, or the I705A substitution within the FH2 domain could rescue actin tail formation in FHOD1-depleted cells. We found that the FHOD1 variant lacking the GBD domain was distributed in the cytoplasm and also in cytoplasmic vesicles in infected cells. This protein did not localize to the formed actin tails and failed to rescue the actin tail formation defect observed in FHOD1-depleted cells (Fig. 5, A and B, FHOD1si, ΔGBD). Expression of the FH1 deletion mutant showed localization to the formed actin tails (Fig. 5, A and B, FHOD1si, ΔFH1). However, this variant failed to rescue actin tail formation (Fig. 5, A and B, FHOD1si, ΔFH1). Finally, the expression of the variant carrying a deletion of the FH2 domain did not localize to the formed actin tails and, as expected, failed to rescue actin tail formation (Fig. 5, A and B, FHOD1si, ΔFH2). Similarly, the I705A substitution within the FH2 domain of FHOD1 failed to rescue actin tail formation. However, we noticed that 60% of the formed actin tails displayed this mutant version of FHOD1 (Fig. 5, A and B, FHOD1si, I705A). Altogether, these results indicate that the functional domains of the formin FHOD1 are required for vaccinia actin tail formation.

Our results indicated a role for the FH1 domain of FHOD1 in vaccinia actin tail formation, suggesting that binding of profilin-actin to the FH1 domain of FHOD1 might be required for FHOD1 function in actin tail formation. To investigate this possibility, we first analyzed the role of Profilin-1 (PFN1) in vaccinia dissemination by using the spreading assay presented in Fig. 1. Using four independent siRNA duplexes targeting PFN1 expression, we found that PFN1 depletion impaired vaccinia dissemination (Fig. 6, A and B, PFN1siA–D). The efficiency of the silencing reagents was confirmed at the transcriptional level (Fig. 6 C). To determine whether the observed spreading defect was related to a defect in actin tail formation, we quantified the number of CEVs displaying actin tails in mock-treated and PFN1-depleted cells. PFN1 depletion resulted in a 2.5-fold decrease in the number of CEVs with actin tails (Fig. 6 D). In addition, we recorded a 30% decrease in the speed of the viruses forming actin tails in PFN1-depleted cells (Fig. 6 E). In support of the specific role of PFN1 in vaccinia actin tail formation, L. monocytogenes displayed wild-type motility in PFN1-depleted cells (Serio et al., 2010; unpublished data). Altogether, these observations suggest that PFN1 specifically contributes to vaccinia dissemination through the initiation and the elongation of vaccinia actin tails.

Our previous results indicated a role for the GBD domain in FHOD1 actin tail localization, suggesting a potential role for Rho/Rac/Cdc42 family members in vaccinia dissemination. Imaging cells expressing GFP-tagged versions of the small GTPases showed that Rac1-GFP localized to the plasma membrane (Fig. 7 A) and was apparently enriched at the membrane surrounding actin tails (Fig. 7 A). In contrast, Cdc42-GFP was very weakly recruited to CEVs and Rho-GFP was not recruited at all (unpublished data). We next quantified the enrichment of Rac1-GFP at the plasma membrane surrounding actin tails. Computer-assisted image analysis of Rac1-GFP signal levels (Fig. 7 A [bottom] and B) showed that membrane projections corresponding to actin tails displayed a 100% increase compared with the signal displayed locally at the plasma membrane (Fig. 7 C). Furthermore, we determined that the p21 binding domain (PBD)–PAK1 probe, which binds activated Rac1 (Hoppe and Swanson, 2004), was also enriched in vaccinia actin tails (Fig. S4). Thus, CEVs trigger signaling events at the plasma membrane that lead to the local recruitment and activation of Rac1.

We also investigated whether Rac1 recruitment relied on the A36–N-WASP–ARP2/3 pathway. To this end we quantified Rac1 enrichment in cells infected with A36-deficient viruses (Fig. 7 D, vΔA36R). Even though vΔA36R viruses did not form actin tails, they displayed active tyrosine kinase signaling and colocalized with phosphorylated Src (Fig. 7 D, αP-Src), as previously reported for the A36R mutant defective for phosphorylation on tyrosine residues 112 and 132 (Newsome et al., 2004). Computer-assisted image analyses (Fig. 7, D [bottom] and E) revealed that the membrane located beneath A36-deficient viruses colocalizing with phospho-Src (P-Src) displayed a 75% increase in dsRed-Rac1 signal levels compared with the levels recorded locally at the plasma membrane (Fig. 7 F). Altogether, these results indicate that signaling events triggered by CEVs at the plasma membrane lead to the A36-independent recruitment of the small GTPase Rac1.

We next determined the functional significance of Rac1 recruitment to the plasma membrane surrounding actin tails. We first used the spreading assay presented in Fig. 1 and quantified the dissemination of vaccinia virus in Rac1-depleted cells. We found that, similar to the situation observed in FHOD1-depleted cells, vaccinia spread from cell to cell was impaired in Rac1-depleted cells (Fig. 8 A). We validated the specific involvement of Rac1 by using four independent silencing reagents targeting Rac1 expression (Fig. 8, A and B, Rac1siA–D). We found that all four siRNA duplexes conferred efficient silencing at the transcriptional level (Fig. 8 C). All four siRNA duplexes also conferred spreading defect phenotypes (Fig. 8, A and B). With the exception of Rac1 siRNA duplex A, treatment with Rac1 siRNA duplexes did not confer general cell growth defect (Fig. S5 A). We also established that Rac1-depleted cells supported L. monocytogenes spread from cell to cell (Fig. S5 B). Similar to the impact of FHOD1 depletion on actin tail formation, we found that Rac1 depletion resulted in a 2.5-fold reduction in the number of CEVs displaying actin tails and a 25% decrease in the speed of vaccinia actin-based motility (Fig. 8, D and E). Altogether, these results indicate that the small GTPase Rac1 is specifically required for vaccinia virus dissemination through stimulation of initiation and elongation of actin tails.

Previous studies have established that Rac1 is specifically involved in FHOD1 activation at the plasma membrane (Westendorf, 2001; Gasteier et al., 2003; Schulte et al., 2008). Our results on vaccinia dissemination indicated that Rac1 and FHOD1 both localized to actin tails and were required for their formation. Altogether these results led us to hypothesize that the recruitment of Rac1 to CEVs mediated the recruitment and activation of FHOD1. To test this hypothesis, we analyzed the localization of FHOD1 in Rac1-depleted cells expressing dsRed-FHOD1. We noticed that the overall level of dsRed-FHOD1 expression was systematically lower in Rac1-depleted cells, suggesting that Rac1 may regulate the overall level of FHOD1 expression through an unknown mechanism. In mock-treated and Rac1-depleted cells expressing comparable levels of dsRed-FHOD1 (Fig. 8 A), we determined that 50% of the actin tails formed in Rac1-depleted cells did not colocalize with FHOD1 (Fig. 8 B), indicating that Rac1 activity contributes to the localization of FHOD1 to actin tails.

As an alternative approach and to confirm the specific role of Rac1 in FHOD1 recruitment, we used a dominant-negative version of Rac1 (Rac1N17). This construct is thought to display its dominant-negative effect through stable interaction with and titration of the guanine nucleotide exchange factors involved in Rac1 activation (Feig, 1999). We compared the effect of expressing dominant-negative versions of Rho, Cdc42, and Rac1 small GTPases on vaccinia actin tail formation. In agreement with previous observations (Arakawa et al., 2007), the number of actin tails formed in cells expressing dominant-negative Rho was similar to the number observed in mock-treated cells (Fig. 9 D, control and DN Rho). In contrast, expression of dominant-negative Rac1 resulted in a threefold decrease in the number of actin tails (Fig. 9, C [top] and D [control and DN Rac1]). Expression of dominant-negative Cdc42 resulted in a twofold decrease in the number of actin tails (Fig. 9 C [top] and D [control and DN Cdc42]). Because previous studies in Cdc42 null cells (Shibata et al., 2002) and Cdc42 depletion by RNAi (unpublished data) indicated that Cdc42 is not required for vaccinia actin tail formation, these results suggested that overexpression of the dominant-negative version of Cdc42 probably affected vaccinia actin tail formation by titrating the guanine nucleotide exchange factor involved in Rac1 activation. Interestingly, we found that overexpression of both full-length FHOD1 or a variant bearing a deletion of the C-terminal DAD domain rescued to wild-type levels the actin tail formation defects observed in cells expressing dominant-negative Rac1 (Fig. 9, C and D, DN Rac1 + FHOD1wt and DN Rac1 + FHOD1ΔDAD). As expected, overexpression of the FHOD1 variant bearing a deletion of the FH2 domain did not rescue the defect observed in cells expressing dominant-negative Rac1 (Fig. 9 D, DN Rac1 + FHOD1ΔFH2). Moreover, overexpression of components involved in the activation of the ARP2/3 complex, such as N-WASP, could not rescue actin tail formation in cells expressing DN Rac1 (Fig. 9 D, DN Rac1 and DN Rac1 + NWASP). Altogether, these results support the notion that Rac1 specifically regulates the activity of FHOD1 in vaccinia actin tail formation.

Our previous results suggested that, in addition to the N-WASP–ARP2/3 pathway, vaccinia virus exploits the Rac1–FHOD1 pathway for efficient actin tail formation. To further characterize the interplay between these two pathways, we quantified the recruitment of N-WASP to CEVs in Rac1-depleted cells. Similar to the results observed in FHOD1-depleted cells (Fig. 4), we found that N-WASP recruitment to CEVs was similar in mock-treated and Rac1-depleted cells (Fig. 10 A, mock and Rac1si). These results therefore indicated that the Rac1–FHOD1 pathway was not required for N-WASP recruitment. Conversely, we tested whether N-WASP may be required for FHOD1 recruitment. To this end, we infected dsRed-FHOD1–expressing cells with wild-type virus expressing a B5R-CFP chimera. We found that >70% of the viruses positive for P-Src recruited FHOD1 in mock-treated cells (Fig. 10, B–D). In contrast, <20% of P-Src–positive viruses recruited FHOD1 in N-WASP–depleted cells (Fig. 10, B–D). Collectively, our observations indicate that the localization of FHOD1 to vaccinia actin tails relies on the small GTPase Rac1 and the nucleation-promoting factor N-WASP.

Although the central role of the ARP2/3 complex and associated regulatory proteins, such as N-WASP, has been demonstrated by unambiguous genetic approaches, it was unclear whether additional cytoskeleton regulators may also play previously unappreciated roles in vaccinia dissemination. Here, we developed high-throughput imaging and computer-assisted image analyses procedures in combination with the RNAi methodology to identify cytoskeleton factors required for vaccinia spread from cell to cell, an approach that led to the identification of a formin, FHOD1. Formin family members harbor a conserved FH2 domain that mediates barbed end activities, including nucleation, capping, and/or elongation of actin filaments. Our results demonstrate that the FH2 domain of FHOD1 plays a critical role in the initiation and the elongation of actin tails, suggesting that FHOD1 supports vaccinia actin tail formation through nucleation and elongation of actin filaments. In further support of the role of FHOD1 through its elongation activity, we show that actin tail formation requires the FH1 domain of FHOD1 and its binding partner PFN1, which stimulates the elongation activity of formins (Romero et al., 2004; Kovar et al., 2006). We note that a recent study reported that, unlike other formin family members, FHOD1 displays capping and bundling activities, but very low elongation and no nucleation activities in vitro (Schönichen et al., 2013). Because our results strongly support the notion that FHOD1 does display nucleation and elongation activities in the context of vaccinia actin tail formation in vivo, we suggest that the apparent absence of nucleation and elongation activities observed in vitro reflects the in vivo complexity of FHOD1 regulation (see Regulation of FHOD1 activity in actin tail formation).

Because FHOD1 depletion did not affect the numbers of CEVs positive for N-WASP, it is unlikely that FHOD1 regulates actin tail formation through regulation of the N-WASP–ARP2/3 pathway. Instead, our results are in agreement with the notion that, in addition to the nucleation activity displayed by the ARP2/3 complex, the nucleation activity of FHOD1 is required for efficient initiation of actin tail formation. Because the ARP2/3 complex requires mother filaments to nucleate daughter filaments, we speculate that FHOD1 primes the process of actin tail formation by nucleating the assembly of actin filaments that constitute mother filaments for the ARP2/3 complex (Fig. 10 E, left). We, however, note that the priming process is probably not absolutely required because tail formation is reduced in FHOD1-depleted cells, but not abolished. Alternatively, the decrease in the numbers of CEVs displaying actin tails may in fact reflect efficient disassembly of tails that do not elongate properly. In addition to a role in regulating actin tail initiation, our results indicate that FHOD1 also plays a role in regulating actin tail elongation. Because formins bind to the barbed ends of actin filaments and increase their rate of elongation (Romero et al., 2004; Kovar et al., 2006), our observations are therefore consistent with the notion that FHOD1 may increase the rate of actin tail elongation by increasing the rate of actin filament elongation after ARP2/3-mediated nucleation of daughter filaments (Fig. 10 E, right). Such a cooperation between the ARP2/3 complex and formins is not unprecedented and was recently described in lamellipodia for ARP2/3 and the formin FMNL2 (Block et al., 2012). In this study, the activity of the formin was regulated by the small GTPase Cdc42 at the plasma membrane and facilitated the elongation of filaments nucleated by the ARP2/3 complex. Thus, the cooperation between the ARP2/3 complex and formins constitutes an emerging theme in actin-based motility, as suggested by our present study on the role of FHOD1 in vaccinia actin tail formation.

In addition to the expected requirement for the FH2 domain, we found that the GBD domain of FHOD1 plays an important role in vaccinia actin tail formation. The interaction between the FH3 and the DAD domains maintains formins in an autoinhibited conformation, which is relieved upon interaction of the GBD domain with activated small GTPases of the Rho family (Li and Higgs, 2003; Lammers et al., 2005). Previous studies determined that FHOD1 preferentially interacts with the small GTPase Rac1 (Gasteier et al., 2003; Schönichen et al., 2006; Schulte et al., 2008). We now provide several lines of evidence supporting the notion that Rac1 mediates vaccinia actin tail formation through the recruitment and activation of FHOD1, independently from the N-WASP–ARP2/3 pathway. We demonstrate that Rac1 is required for FHOD1 recruitment (Fig. 9), but not for N-WASP recruitment (Fig. 10). Moreover, we show that the actin tail formation defects observed in cells expressing dominant-negative Rac1 are specifically rescued by overexpression of FHOD1, but not by overexpression of a mutant version of FHOD1 lacking the FH2 domain or by overexpression of the wild-type version of N-WASP. Although Rac1 is apparently not required for N-WASP recruitment, N-WASP is nonetheless required for FHOD1 recruitment (Fig. 10 D). We speculate that adaptor proteins such as WISH, which binds FHOD1 (Westendorf and Koka, 2004) and N-WASP (Fukuoka et al., 2001), may contribute to the N-WASP–dependent recruitment of FHOD1.

Because our results indicated that Rac1 was enriched and activated at the plasma membrane surrounding vaccinia actin tails, we tested the possibility that, similar to N-WASP, the recruitment and activation of Rac1 may rely on the viral protein A36. However, P-Src–positive, A36-deficient CEVs recruited Rac1 as efficiently as wild-type viruses, indicating that Rac1 recruitment did not rely on the viral protein A36. Thus, the signaling events leading to the recruitment and activation of Rac1 beneath CEVs remain to be determined. We speculate that, in addition to mediating A36 phosphorylation, Src-mediated tyrosine kinase signaling mediates Rac1 activation, which in turn leads to the recruitment and activation of FHOD1 (Fig. 10 E). In agreement with this notion, Src was shown to regulate the activity of FHOD1 in unrelated systems (Koka et al., 2005; Hannemann et al., 2008).

In conclusion, we propose that vaccinia virus evolved the ability to display robust actin-based motility by manipulating not only the N-WASP–mediated activation of the ARP2/3 complex but also the Rac1-mediated activation of FHOD1 (Fig. 10 E).

HeLa (American Type Culture Collection), HeLa S3 (American Type Culture Collection), and BS-C-1 (American Type Culture Collection) cells were grown in DMEM (Invitrogen) supplemented with 10% FBS (Gibco) at 37°C in a 5% CO₂ incubator. Vaccinia virus strains WR, vB5R-GFP (Ward and Moss, 2001b), B5R-GFP/deltaA36R (Ward and Moss, 2001a), and the B5R deletion mutant vSI-14 (Wolffe et al., 1993) were provided by B. Moss (National Institute of Allergy and Infectious Diseases, Bethesda, MD). The recombinant vA5L-mCherry/3xF-B5R virus carrying an N-terminal fusion of A5L to mCherry and a triple FLAG-tagged version of B5R was previously described (Alvarez and Agaisse, 2012). Viral strains were grown and propagated in HeLa cells and plaque assays were performed in BS-C-1 cells according to standard procedures (Earl et al., 2001). In brief, 6-well plates were seeded with 5 × 10⁵ BS-C-1 cells per well and infected on the next day with 10-fold serial dilutions of the virus stocks prepared in DMEM supplemented with 2.5% FBS. 2 d after infection cells were stained with 0.1% crystal violet and the virus titer was determined by plaque counting. Virus stocks were prepared from HeLa S3 cells and partially purified by centrifugation through a 36% sucrose cushion (Earl et al., 2001).

The procedure for siRNA setup in 384-well plates was described previously (Chong et al., 2011). In brief, 10 µl of individual siRNAs (200 nM) were aliquoted in 384-well plates. 10 µl of Dharmafect 1 transfection mixture (5 µl of Dharmafect 1 per 1 ml of serum-free DMEM) was dispensed per well and incubated for 20 min at room temperature. HeLa cells were lifted and resuspended at 100,000 cells/ml in DMEM supplemented with 10% FBS. 20 µl of the cell suspension were delivered to each well, and the plates were then incubated for 3 d at 37°C, in 5% CO₂. siRNA-treated cells were infected at 0.01 pfu per cell with 10 µl of a dilution of vaccinia virus vB5R-GFP stock. The plates were centrifuged at 2,000 rpm for 10 min and infection was allowed to proceed for 16 h. Infected cells were fixed in 4% paraformaldehyde in PBS and then stained with Hoechst 33348 (Molecular Probes), washed in PBS, and imaged.

384-well plates were imaged using a microscope (TE 2000; Nikon) equipped with a digital CCD camera (Orca ER; Hamamatsu Photonics), motorized stage (Prior), motorized filter wheels (Sutter Instrument), and a 10× objective (NA 0.45; Nikon) mounted on a piezo focus drive system (Physik Instrumente). Image acquisition and analyses were conducted using the MetaMorph 7.1 software (Molecular Devices). Appropriate threshold of green channel identified objects corresponding to vB5R-GFP–infected cells (Fig. S1 B, threshold). We next used the image morphometry analysis module and the dilate function in the morphology filters to generate a mask displaying foci of infection (Fig. S1 B, segmentation). This mask was used as the source image to count and classify foci of infection according to their size into large and small foci (Fig. S1 B, classifier). The spreading index was calculated as the ratio [total foci − small foci]/[total foci] for a given well.

A plasmid containing B5R fused to CFP and ∼500 bp of flanking sequence on each side was constructed by excision of the ECFP ORF from pECFP-N1 (Takara Bio Inc.) with NcoI and NotI restriction enzymes and subcloning into pBMW10 (Ward and Moss, 2001b), provided by B. Ward (University of Rochester, Rochester, NY). The B5R-CFP fusion plasmid was transfected with X-tremeGENE 9 (Roche) into HeLa cells that had been infected with the vaccinia virus B5R deletion mutant vSI-14 (Wolffe et al., 1993). Recombinant viruses that formed cyan fluorescent foci were plaque purified three times. The resulting recombinant vB5R-CFP was analyzed by PCR. Primers (13) and (16) amplified a 2.3-kb DNA product in agreement with the insertion of the B5R-CFP ORF.

Cells were transfected by reverse transfection with Dharmafect1 (Thermo Fisher Scientific) and individual siRNAs (50 nM final) or a pool of the four silencing reagents (12.5 nM each, 50 nM total) and were incubated for 72 h in a 24-well plate format. The sequences targeted by the individual siRNA duplexes are listed in 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, with the addition of DNaseI for the removal of unwanted genomic DNA. mRNA levels were determined by quantitative real-time PCR using the LightCycler 480 Master kit and LightCycler 480 instrument (Roche). The probes and primers were designed according to Roche recommendation.

For Western blot analysis, cells were transfected with the corresponding silencing reagents in a 24-well plate format and lysed directly in Laemmli sample buffer. The anti-FHOD1 antibody (1:1,000) was obtained from Novus Biologicals. The anti-actin antibody (1:10,000) was obtained from Sigma-Aldrich. The anti-rabbit HRP-conjugated secondary antibody (1:10,000) was obtained from Jackson ImmunoResearch Laboratories, Inc.

HeLa cells were seeded on glass coverslips, grown to 30–50% confluency, and infected with vaccinia virus at 10 pfu per cell. For vaccinia infection, cells were incubated with the virus inoculum for 1 h, and then washed three times with PBS before adding fresh media. At 10 h after infection cells were washed with PBS, fixed in ice-cold 3% paraformaldehyde in PBS, and washed twice with PBS. To detect CEVs at the cell surface, cells infected with vA5L-mCherry/3xF-B5R were fixed and then blocked in 10% BSA/PBS, stained in 2% BSA/PBS containing mouse M2 monoclonal antibody anti-FLAG (1:1,000; Sigma-Aldrich) for 30 min, and washed five times in 2% BSA in PBS before addition of secondary antibody. To visualize actin tails, cells were permeabilized in 0.1% Triton X-100 in PBS and stained with phalloidin. Actin tails were identified as bright objects in the phalloidin staining and were required to be at least 2 µm long. To quantify localization of N-WASP to CEVs, cells were first stained for surface FLAG, then blocked and permeabilized in 10% BSA/0.1% Triton X-100 in PBS, and stained with anti–N-WASP rabbit mAb (1:200; Cell Signaling Technology). Rabbit polyclonal antibody against A36RC (1:1,000) was provided by B. Moss. P-Src was stained with a mouse monoclonal antibody anti-P-Src (Tyr416; 1:200) obtained from EMD Millipore. Alexa Fluor secondary antibodies (1:1,000) and Alexa Fluor–conjugated phalloidin (1:500) were obtained from Invitrogen. Coverslips were mounted onto glass slides using FluoroGuard Antifade Reagent (Bio-Rad Laboratories) and imaged using a TE 2000 microscope (Nikon) equipped with a digital CCD camera (Orca ER), motorized stage, motorized filter wheels, and a 60× objective (NA 0.85; Nikon) mounted on a piezo focus drive system. Image acquisition and analysis was conducted using the MetaMorph 7.1 software. Quantitative analysis of CEV formation in cells infected with the recombinant vA5L-mCherry/3xF-B5R virus was conducted as previously described (Alvarez and Agaisse, 2012). In brief, appropriate threshold and size filtering of red channel to identify objects corresponding to virus cores expressing A5-mCherry generated a mask displaying only single virus particles. The same analysis was performed for the FLAG staining channel to identify objects corresponding to FLAG-B5–expressing CEVs. Finally we used logical AND in the arithmetic function to define and count CEVs as objects corresponding to colocalization of A5-mCherry and FLAG-B5 signals. The quantification of percentage localization with CEVs was performed by visual inspection.

The pCMV5-HA-FHOD1 construct was a gift from O. Fackler (University Hospital Heidelberg, Heidelberg, Germany; Madrid et al., 2005). Full-length FHOD1 was amplified from pCMV5-HA-FHOD1 using primers (18) and (19) and cloned into pEGFP-C1 (Takara Bio Inc.) between EcoRI and BamHI restriction sites. To generate an siRNA-resistant form of FHOD1, a PCR fragment containing silent mutations in the target sequence for FHOD1 siRNA B (Table S1) was generated by fusion of the PCR products amplified with primers EFGP C and (31) and primers (32) and (19) using pEGFP-FHOD1 as a template. This was cloned into pdsRed-C1 between XhoI and BamHI restriction sites to yield siRNA-resistant pdsRed-FHOD1. The corresponding GBD domain deletion construct was obtained by cloning the PCR product amplified with primers (58) and FHOD1rev between XhoI and HindIII restriction sites. To construct the FH1 domain deletion, the PCR products obtained with primers (32) and (55) and primers (54) and (19) were fused together and cloned into pdsRed-FHOD1 between HindIII and BamHI restriction sites. The FH2 domain deletion was generated following the same cloning strategy after fusing together the PCR products generated with primers (32) and (36) and primers (35) and (19). Similarly, the PCR products obtained using primers (32) and (71) and primers (70) and (19) were fused together to generate the Ile to Ala substitution at amino acid position 705. The deletion of amino acids 1005–1064 comprising the DAD domain was generated using primers (32) and (73).

The wild-type and dominant-negative versions of Rho, Rac, and Cdc42 cDNA were described previously (Ridley et al., 1992; Bagrodia et al., 1995; Qiu et al., 1995). Dominant-negative RhoN19, RacN17, and Cdc42N17 small GTPases carry a Thr to Asn substitution at the indicated amino acid position. The corresponding GFP fusions were expressed from pEGFP-C1 vector and were provided by J. Galán (Yale University, New Haven, CT). The dsRed-Rac1 construct was obtained by subcloning the BspEI–BamHI fragment from pEGFPC1-Rac1 (Krall et al., 2002) into pDsRed-Monomer-C1. The PBD of PAK1 was amplified by PCR using primers (66) and (67) and cloned into pdsRed-N1 between XhoI and HindIII restriction sites. N-WASP ORF was amplified from HeLa cell cDNA with primers (37) and (38) and cloned into pYFP-C1 between XhoI and EcoRI. Actin-CFP was derived from actin-GFP plasmid (Rolls et al., 1999). The oligonucleotide sequences of the primers used in this study are listed in Table S2.

Mock- and siRNA-treated HeLa cells were seeded on glass coverslips in 24-well format on day 0. DNA constructs were transfected with X-tremeGENE 9 (Roche) on day 2. On day 3, cells were infected with vaccinia virus as described for immunofluorescence procedures, fixed, and processed for imaging using a spinning disc confocal microscope (TE2000E; Nikon).

Cells expressing Rac1-GFP were infected with WR strain of vaccinia virus. At 10 h after infection cells were fixed and stained with phalloidin. Images were acquired on a spinning disc confocal microscope (TE2000E) and a 60× oil objective (NA 1.4) using the Volocity software (PerkinElmer). Rac1 enrichment in actin tails was quantified on XZ reconstructions of the confocal images using the line profile measurement module of the Volocity software. The Rac1 enrichment in actin tails represents the ratio between the maximum intensity value in the Rac1-GFP channel obtained from a line crossing a given actin tail and the maximum value from a line crossing the plasma membrane adjacent to the actin tail (see Fig. 6). PBD-PAK1 intensity was measured in cells coexpressing PBD-PAK1-DsRed and Rac1-GFP. The PBD-PAK1 enrichment at a given vaccinia tail or filopodium base and tip was quantified in the PBD-PAK1-DsRed channel and represents the ratio between the intensity value for PBD-PAK1 at the maximum of Rac1 intensity measured in the blue channel and the corresponding value at the adjacent plasma membrane. Rac1 enrichment for the ΔA36R virus variant was performed on samples stained for P-Src with mouse monoclonal antibody anti-P-Src (Tyr416; 1:200). Rac1 enrichment was quantified in the Rac1-DsRed channel on P-Src–positive viral particles, as described for wild-type virus.

HeLa cells were seeded on 35-mm imaging dishes (MatTek Corporation). Mock- and siRNA-treated cells were cotransfected with pYFP–N-WASP and actin CFP DNA constructs the day before infection with WR strain of vaccinia virus. Images were captured at 8–10 h after infection every 12 s on a spinning disc confocal microscope (TE2000E) using a 60× oil objective (NA 1.4) and the Volocity software. Tracking of N-WASP–tipped actin tails was performed using the tracking module of the Volocity software. N-WASP–tipped actin tails tracked to calculate mean velocity were required to be motile for at least 60 s.

Quantitative analyses were performed in three independent experiments and 30–50 cells from 20 independent images were counted for each condition. Quantifications on single viruses were performed on at least 500 objects. Data are presented as mean ± SEM) and were analyzed by Student’s t test using Prism 5 (GraphPad Software).

Fig. S1 shows a time course of infection of vaccinia in control and ARPC4-depleted cells, and the computer-assisted image analysis of spreading. Fig. S2 shows that FHOD1-depleted cells support L. monocytogenes actin-dependent dissemination. Fig. S3 shows the specificity of the recruitment of FHOD1 to vaccinia actin tails. Fig. S4 shows the enrichment of PBD-PAK1 probe in vaccinia actin tails. Fig. S5 shows that Rac1-depleted cells support L. monocytogenes actin-dependent dissemination. Table S1 shows the sequences targeted by individual siRNA duplexes against FHOD1, RAC1, and PFN1. Table S2 shows the sequences of the oligonucleotides used in this study.

We thank Sebastian Hannemann and the members of the Agaisse Laboratory for critical discussion of the manuscript.

This work was supported by the Pew Latin American Fellows Program in the Biomedical Sciences (D.E. Alvarez) and National Institutes of Health grant R21AI107150 (H. Agaisse).