Bim and Bmf Synergize To Induce Apoptosis in Neisseria Gonorrhoeae Infection

Bcl-2 family proteins including the pro-apoptotic BH3-only proteins are central regulators of apoptotic cell death. Here we show by a focused siRNA miniscreen that the synergistic action of the BH3-only proteins Bim and Bmf is required for apoptosis induced by infection with Neisseria gonorrhoeae (Ngo). While Bim and Bmf were associated with the cytoskeleton of healthy cells, they both were released upon Ngo infection. Loss of Bim and Bmf from the cytoskeleton fraction required the activation of Jun-N-terminal kinase-1 (JNK-1), which in turn depended on Rac-1. Depletion and inhibition of Rac-1, JNK-1, Bim, or Bmf prevented the activation of Bak and Bax and the subsequent activation of caspases. Apoptosis could be reconstituted in Bim-depleted and Bmf-depleted cells by additional silencing of antiapoptotic Mcl-1 and Bcl-XL, respectively. Our data indicate a synergistic role for both cytoskeletal-associated BH3-only proteins, Bim, and Bmf, in an apoptotic pathway leading to the clearance of Ngo-infected cells.


Introduction
Infection with various pathogens results in the inhibition or activation of apoptotic cell death [1]. Whereas viral pathogens frequently inhibit host cell apoptosis, many bacteria kill immune or epithelial cells by apoptosis allowing them to subvert immune reactions or to invade tissues, respectively. The obligate human specific bacterium Neisseria gonorrhoeae (Ngo), the causative agent of the sexually transmitted disease gonorrhea, induces apoptosis in genital epithelia. Since induction of apoptosis requires the firm attachment of the gonococci to host cells [2], exfoliation of infected epithelial cells covered with adherent bacteria has been suggested as the immediate cellular responses against infection [3,4]. This detachment-associated apoptosis of infected cells resembles anoikis, a special form of apoptosis that is induced by absent or inappropriate cell-matrix interactions [5].
Bcl-2 family proteins control mitochondrial outer membrane permeabilization (MOMP), which is the critical step in many forms of apoptosis [6,7]. The Bcl-2 family consists of pro-and antiapoptotic members that share homologies within their Bcl-2 homology domains (BH). The antiapoptotic Bcl-2 family proteins harbor BH1-4 domains and presumably act by sequestering and inhibiting proapoptotic Bcl-2 members [8]. Proapoptotic Bcl-2 family proteins can be further subdivided into the branch of pore forming, multidomain BH1-3 proteins (like Bak and Bax) and the BH3-only branch (including Bim, Bmf, Bid, Bad, Noxa and Puma) [9,10]. Active BH3-only proteins cause conformational changes within Bak and Bax, which subsequently homooligomerize and form pores in the outer mitochondrial membrane [11,12]. MOMP culminates in the release of proapoptotic proteins like cytochrome c, leading to the activation of caspases and caspase-independent death effectors [13].
The mechanisms through which BH3-only proteins activate Bak or Bax are not fully understood. BH3-only proteins may release Bak and Bax from inhibition by anti-apoptotic Bcl-2 protein [14]. Alternatively, the group of BH3-only proteins may include two subgroups, namely survival antagonists that neutralize BH1-4 proteins, and death agonists that activate BH1-3 proteins [15,16]. A competition of death agonists and survival antagonists for the binding to BH1-4 proteins has been reported [17]. Upon binding of survival antagonists, death agonists are released from their sequestration by BH1-4 proteins and hence freed to act directly on BH1-3 proteins.
Cytotoxic stimuli activate BH3-only proteins by a variety of distinct mechanisms such as p53-dependent transcriptional regulation (Puma and Noxa [18,19]), proteolytic cleavage (Bid [20]), dephosphorylation (BAD [21]) or phosphorylation (Bim and Bmf). Under normal circumstances, Bim and Bmf are sequestered via dynein light chains (DLC) to the actin and tubulin cytoskeleton, respectively, which prevents them from activating Bak and Bax [22,23]. Previous work has suggested that phosphorylation of Bim and Bmf within their DLC-binding sites is mediated by Jun-N-terminal kinase-1 (JNK-1) facilitating the release of both proteins from the cytoskeleton during anoikis [24,25].
Here, we analyzed the signaling pathways upstream of Bak and Bax in Ngo-infected cells. Unexpectedly, both Bim and Bmf were found to act in concert to induce apoptosis of Ngo infected cells. Our data suggest a role of Ngo infection-induced cytoskeletal reorganization in the initiation of apoptosis pathways.

Caspase-independent detachment of infected cells
Exfoliation of epithelial cells has previously been described to be caused by Ngo infection [3,4,26]. Since this process resembles anoikis, we further investigated the connection between Ngoinduced cell detachment and apoptosis. HeLa cells were infected with Ngo VPI (N242), a clinical isolate and morphological changes were correlated with the activation of caspases. Detachment from the culture support was visible as soon as 6 to 9 h post-infection ( Figure 1A) concomitant with the proteolytic maturation of caspase 3 ( Figure 1B). To test whether caspase activity is required for exfoliation, cells were infected in the absence or presence of the pan-caspase inhibitor Z-VAD-fmk and then were analyzed by electron and fluorescence microscopy. In the presence of Z-VADfmk, Ngo-infected cells continued to detach yet remained otherwise intact and hence failed to disintegrate by apoptosis ( Figure 1C and 1D, Video S1) while the activation of caspases 3 and 7 was blocked during the entire duration of the experiment ( Figure 1E and 1F). Detachment was further analyzed by acquiring z stacks of infected cells by laser scanning confocal microscopy and subsequent 3-dimensional remodeling ( Figure 1G and Video S2). The detachment and induction of apoptosis is not a general response of these cells to infection stress since the same cell line exhibits marked apoptosis resistance as consequence of infection with C. trachomatis [27,28]. These results demonstrate that caspases are required for the apoptotic disassembly of Ngoinfected cells but not for their detachment.
To test whether the observed effect is specific for HeLa cells and the bacterial strain VPI (N242), we tested other gonococcal derivatives for their capacity to induce exfoliation and cell death. Eleven clinical gonococcal isolates from different patients isolated from blood, urethra, cervix, vagina or urine were analyzed. The capacity of these strains to adhere to HeLa cells correlated well with the induction of exfoliation and apoptosis (Table 1; Figure  S1), suggesting that exfoliation and apoptosis induction is a common effect of adherence to HeLa cells. We also investigated whether the effect is specific for HeLa cells. N242 induced significant apoptosis in ME180 and Hep2 cells. From the genetically defined derivatives of the laboratory strain MS11, only strain N920 which forms pili induced apoptosis in ME180 cells, whereas strains expressing Opa57 (N1163, not shown) or Opa52 (N309) failed to induce significant apoptosis in HeLa, Hep2 and ME180 cells ( Figure S2). N242 is a clinical isolate expressing 5 different Opacity-associated (Opa) proteins required for the binding of these bacteria to heparane sulfate proteoglycanes (HSPG), integrins or 'carcinoembryonic antigen-related cell adhesion molecules' (CEACAMs) of the host cell ( [29]; for review see [30]). We next tested whether laboratory strains gain the capacity to induce detachment of HeLa cells expressing Opa protein receptors. The MS11 strain MS11 N927 (Opa 2 ; PorB IA ) failed to induce detachment of HeLa cells permanently expressing CEACAM 1 or CEACAM 3 respectively ( Figure S3) [31]. In contrast, the MS11 strain N1163 (Opa57; PorB IA ) induced detachment of infected cells in the CEACAM 1 expressing HeLa cells but not in the cell line expressing CEACAM 3 ( Figure S3). These data suggested an interaction-, and in addition a receptor-specific mechanism underlying detachment and apoptosis induction.

Synergistic action of BH3-only proteins during apoptosis
Since infection of HeLa cells with N242 caused the most prominent effects, we focused on this system to further investigate the mechanisms underlying infection-induced apoptosis. We have previously demonstrated that infection with Neisseria induced the activation of Bak and Bax and finally apoptotic cell death [32]. To delineate the signaling pathway leading to the activation of Bak and Bax, we systematically depleted BH3-only proteins in a RNA interference miniscreen. The knockdown of the siRNAs was validated by quantitative real-time PCR (.75% knockdown at the mRNA level) and immunoblot analysis (.75% knockdown at the protein level) (Figure 2A and 2B). Knockdown of Bim and Bmf (but not that of Bid or Bad) resulted in a significant reduction of effector caspase activity, as measured with a fluorogenic caspase 3/ 7 substrate or by immunochemical detection of proteolytically mature caspase 3 ( Figure 2C and 2D, and Figure S4). Bim and Bmf knockdown specifically inhibited the caspase activation induced by Ngo ( Figure 2C and 2D, and Figure S5), yet had no effect on caspase activation induced by the genotoxic agent cisplatin ( Figure S5A). Ngo infection failed to induce Puma, Noxa and any of the tested mRNAs for BH3 only proteins ( Figure 2E, and Figure S6), although cisplatin was able to activate the transcription of both the Puma and Noxa genes ( Figure S5B and Figure S5C). Accordingly, the depletion of Puma or Noxa did not affect caspase-3 activation in Ngo-infected cells ( Figure 2F). These data demonstrate that Bim and Bmf are specifically required for the Ngo-triggered activation of caspases.

Release of Bim and Bmf from the cytoskeleton during infection
In healthy cells, Bim and Bmf are sequestered to the cytoskeleton by binding to dynein light chains. In response to cytotoxic stimuli, that induce cytoskeletal rearrangements, Bim

Author Summary
A variety of physiological death signals, as well as pathological insults, trigger apoptosis, a genetically programmed form of cell death. Pathogens often induce host cell apoptosis to establish a successful infection. Neisseria gonorrhoeae (Ngo), the etiological agent of the sexually transmitted disease gonorrhoea, is a highly adapted obligate human-specific pathogen and has been shown to induce apoptosis in infected cells. Here we unveil the molecular mechanisms leading to apoptosis of infected cells. We show that Ngo-mediated apoptosis requires a special subset of proapoptotic proteins from the group of BH3-only proteins. BH3-only proteins act as stress sensors to translate toxic environmental signals to the initiation of apoptosis. In a siRNA-based miniscreen, we found Bim and Bmf, BH3-only proteins associated with the cytoskeleton, necessary to induce host cell apoptosis upon infection. Bim and Bmf inactivated different inhibitors of apoptosis and thereby induced cell death in response to infection. Our data unveil a novel pathway of infectioninduced apoptosis that enhances our understanding of the mechanism by which BH3-only proteins control apoptotic cell death.
and Bmf may act as stress sensors in thus far that they are released from the cytoskeleton and induce the activation of Bak and Bax [22,23]. Accordingly, cytoskeleton fractions obtained from Ngoinfected cells generally contained less Bim and Bmf than those from non-infected control cells ( Figure 2G and Figure S7A). A similar result was obtained when cytoskeleton and cytosol were separated by sucrose gradient centrifugations. Bim and Bmf from infected samples shifted from heavier cytoskeleton containing to lighter fractions ( Figure S7B), indicating a release of these proteins from the cytoskeleton in infected cells. Addition of Z-VAD-fmk did not prevent the release of Bim and Bmf from the cytoskeleton ( Figure 2G), demonstrating that this phenomenon occurs independently of caspase activity.

JNK-dependent activation of Bim and Bmf
Active JNK-1 is reportedly sufficient for the release of Bim and Bmf from the cytoskeleton [24]. We have previously shown that JNK-1 is activated already 30 minutes after infection with Neisseria, leading to NFkB activation and proinflammatory responses [33].
Although the short-term effects of JNK-1 activation can be cytoprotective, prolonged JNK activation induces apoptotic cell death [34]. Phosphorylated, active JNK-1 could be detected for the whole period of Ngo infection up to 15 h ( Figure 3A), correlating with a reduced electrophoretic mobility of Bim and Bmf at later timepoints ( Figure 3B and Figure S8). ERK seems not to be involved in the signaling as there was no activation upon infection ( Figure S9). Silencing of JNK-1 with validated siRNAs ( Figure 3C) prevented the shift in the size of Bim and Bmf ( Figure 3B, 3C, and 3D), indicating a role of JNK-1 in posttranslational modification of Bim-L and Bmf in Ngo-infected cells.
The caspase activity of Ngo-infected cells depleted of JNK was reduced to the same level as that of cells subjected to the knockdown of Bim or Bmf ( Figure 3E). Moreover, the frequency of cells with apoptotic chromatin condensation was reduced In JNK-1-depleted as compared to control cells ( Figure 3F). JNK-1 depletion also partially inhibited the Ngo-induced release of Bim and Bmf from the cytoskeleton ( Figure 3G). In addition inhibiting JNK-1 by means of a chemical inhibitor partially reduced a translocation of Bim and Bmf from heavier to lighter fractions in sucrose gradients ( Figure S7C). In conclusion, JNK-1 depletion can prevent the post-transcriptional modification of Bim and Bmf, reduce their loss from the cytoskeleton fraction and inhibit apoptosis of Ngo-infected cells.

Rac required for cytoskeletal and apoptotic signaling
The Rho-GTPases are central regulators of cytoskeletal changes initiated by extracellular signals. Most prominent, Rho and Rac have been shown to be involved in neisserial uptake and phagocytosis [35]. Therefore, we reanalyzed the link between Rac and JNK signaling during Ngo-induced apoptosis [36]. The knockdown of Rac-1 or its inhibition with the pharmacological agent NSC23766 abolished the reorganization of the cytoskeleton initiated by Ngo infection ( Figure 4A). Rac inhibition caused a significant reduction in caspase activation, apoptotic nuclear fragmentation ( Figure 4B and 4D), JNK-1 phosphorylation ( Figure 4D) and Bim and Bmf cytoskeletal release ( Figure S7D). These experiment place Rac upstream of JNK-1 and all JNK-1dependent apoptotic events affecting Ngo-infected cells.

Bim and Bmf triggered activation of Bak and Bax
The role of Bim and Bmf in the activation of Bak and Baxdirect activation or deinhibition? -is still a matter of controversy [8,15,17]. Irrespective of their exact of mode of action the function of Bim and Bmf in Ngo-induced apoptosis cannot be redundant because depletion of either of them prevented the induction of apoptosis by Ngo infection. To unravel the mechanisms of this non-redundancy, we assessed the activation of Bak and Bax by means of antibodies that recognize their exposed N-termini and hence their activated conformation. SiRNA-and shRNA-mediated knockdown of Bim as well as siRNA-mediated knockdown of Bmf prior to Ngo infection prevented the activation of both Bak and Bax ( Figure 5A and 5B and Figure S4), underlining the essential need of both Bim and Bmf in this pathway.
In certain apoptosis pathways like Ngo infection or cisplatin induction, Bak and Bax become activated in a hierarchical manner, with Bak acting upstream of Bax [32,37]. The activation of Bak involves its release from antiapoptoticBcl-2 analogues such as Mcl-1 [38]. Combined silencing of Mcl-1 and Bim, but not that of Mcl-1 and Bmf or Mcl-1 knockdown alone reestablished the apoptotic program triggered by Ngo infection (Figure 5C), suggesting that Bim acts as a specific Mcl-1 antagonist in this system. Combined silencing of Bcl-X L plus Bmf, but not that of Bcl-X L and Bim or Bcl-X L alone also reestablished Ngo-induced apoptosis ( Figure 5C). In contrast, Bcl-2 co-silencing had no apoptosis-sensitizing effect on either Bim-or Bmf-depleted cells. Potential off target effects within the same protein family could be excluded by systematic cross analysis. In particular, Bim-and Bmf silencing did not cause deregulated expression of anti-apoptotic members of the Bcl-2 family ( Figure S10), ruling out an indirect effect of Bim and Bmf depletion by overexpression of apoptosis inhibitors. We concluded from these data that Bim and Bmf activate apoptotic pathways by functionally sequestering Mcl-1 and Bcl-X L , respectively.

Discussion
Neisseria gonorrhoeae is a highly adapted human pathogen that utilizes multiple adhesins to interact with host cell receptors to trigger cytoskeletal reorganization, invasion or phagocytic uptake, intraphagosomal accommodation, nuclear reprogramming of host cells, cytokine/chemokine release and finally host cell apoptosis [39]. By investigating the apoptotic pathway involved in the infection-induced activation of Bak and Bax, we discovered an unexpected connection between pathogen-induced cytoskeletal reorganization and apoptosis. Attachment of bacteria initiated the activation of Rac-1 leading to rearrangement of the cytoskeleton (which is presumably required for exfoliation) and the activating phosphorylation of the stress kinase JNK-1. JNK-1 then participated in the activation of the BH3-only proteins Bim and Bmf that together facilitate Bak-and Bax-dependent apoptosis.
Besides the well characterized isolate N242 [29], several other clinical isolates induced exfoliation and apoptosis indicating that gonococci trigger similar pathways leading to cell death. Our preliminary data on the initial trigger of cell detachment leading to cell death unveiled a role of specific adhesin -receptor interactions. N242 induced exfoliation and cell death in different cell lines tested. These effects very likely depend on the interaction of one or more of the expressed Opa proteins with a yet uncharacterized receptor. Although derivatives of strain MS11 failed to induce apoptosis in HeLa cells, a similar efficient response as with N242 was observed with derivative N1163 (Opa57; PorB IA ) upon infection of HeLa-CEACAM1 but not in HeLa-CEACAM3. Interestingly, CEACAM-1 has been shown to be upregulated in primary ovarian surface epithelial cells by gonococcal infection suggesting that the interaction with this receptor has in vivo relevance [40]. Moreover, the specificity for one CEACAM-recombinant cell line over the other is interesting, because both have been demonstrated to be susceptible for infection with Opa57-expressing gonococci [41]. It is therefore likely that particular adhesin-receptor interactions determine the detachment and apoptosis induction as consequence of this cellpathogen interaction. This assumption would be in agreement with several reports on the inhibition of apoptosis by gonococcal infection [42]. In one of these studies, Bim was downregulated upon infection of epithelial cells with a piliated gonococcal derivative [43], supporting a central role of Bim in life-death decisions as consequence of gonococcal infections.
Numerous bacterial pathogens induce the reorganization of the host cell cytoskeleton, often initiating the active uptake of bacteria  [44]. Nevertheless, the activation of apoptosis is not a common outcome of such bacterial infections. Our data suggest that downstream of cytoskeletal reorganization, the prolonged activation of JNK-1 is required for lethal signaling. It is interesting to note that short term activation of JNK induces antiapoptotic and proinflammatory responses in the host cell infected with Ngo [33,45]. JNK may therefore exert a dual function during Ngo infection, first by protecting the cell for a short period postinfection and then by triggering the exfoliation of the infected cells. We show here that JNK was required for Bim-and Bmfdependent apoptosis during infection, consistent with the previously described JNK-specific phosphorylation of Bim and Bmf within their dynein binding domains [24]. Accordingly, the release of Bim and Bmf from the cytoskeleton as well as their reduced electrophoretic mobility was reduced in JNK-1-depleted cells ( Figure 3G).
The exact mode of BH3-only activity is still being discussed. Here we show that both Bim and Bmf are essential to induce Bak and Bax activity for Ngo-triggered apoptosis. As the double knockdown of Bim and Mcl-1 re-sensitized cells for apoptosis, the action of Bmf alone seems to trigger apoptosis efficiently in the absence of Mcl-1. Likewise, apoptosis could be rescued in the absence of Bmf by co-knockdown of Bcl-X L , suggesting that the action of Bim alone suffices to induced apoptosis in the absence of Bcl-X L . In this scenario, both Bim and Bmf need to be activated for efficient induction of cell death due to their joint capacity to inhibit two different anti-apoptotic Bcl-2 homologues (see model in Figure 6). As a result, this study furnishes yet another example for the complex relationship between antagonizing pro-and antiapoptotic Bcl-2 family proteins.

Cell culture, bacterial strains, and infection
HeLa cells (human cervix carcinoma, later diagnosed as adenocarcinoma) ATCC CCL2 and HeLa cell lines expressing recombinant CEACAM receptors [31] were grown in RPMI 1640 (Gibco) supplemented with 10% heat inactivated FCS in the presence of 5% CO 2 . The cells were routinely passaged every 2-3 days and the passage number never exceeded 20 passages before a new batch with a low passage number was used. Cells were seeded 24 h before infection and were washed several times with RPMI before infection. HEK 293T cells (immortalized human embryonic kidney) ATCC CRL-11268 for the production of virus and HEp-2 (ATCC CCL-23) were grown in DMEM (Gibco) supplemented with 10% heat inactivated FCS. and ME-180 (HTB-33) were grown in McCoys 5a supplemented with 10% heat inactivated FCS respectively under the same conditions.
The following N. gonorrhoeae strains and derivatives were used in this study: The clinical isolate Ngo strain VP1 (N242; PorB IA ; P 2 ; Opa 27 ; Opa 27,5 ;Opa 28 ; Opa 29 ; Opa 30 ; LPS type L1) [29]; Ngo strain MS11 derivatives N302 (PorB IB ; P 2 ; Opa 2 ), N920 (PorB IA ; P + ; Opa 2 ), a PorB IA derivative of N917 (PorB IB ; P + ; Opa 2 ) and N927 (PorB IA ; P 2 ; Opa 2 ) have been described [46,47]. MS11 derivative N1163 (PorB IA ; P 2 ; Opa 57 ) is a PorB IA derivative of  strain N313 [41]. Clinical gonococcal isolates from Germany were obtained anonymously from the strain collection of the National Reference Laboratory for Meningococci hosted by the Institute for Hygiene and Microbiology at the University of Würzburg. Species confirmation for those strains was obtained at the Reference Laboratory by standard biochemical tests and partial 16S rRNA sequencing. Gonococci were grown on GC agar base plates (Becton Dickinson, Difco and Remel) supplemented with Proteose Pepton Nr. 3 (Difco) and 1% vitamin mix for 14-20 h at 37uC in 5% CO 2 in a humidified atmosphere. Infections were routinely performed in the absence of FCS at a multiplicity of infection (MOI) of 1. If not indicated else wise the respective assays were carried out after 15 h of infection with N242.
For the inhibition of caspases, cells were pre-incubated with 50 mM Z-VAD-fmk (Bachem) for 15 min prior to infection and throughout the experiment. Cisplatin was used at a concentration of 50 mM for 20 h in supplemented media.

FACS assays
Caspase 3 and 7 activities were measured by CaspACE assay. Control and infected cells were collected and stained with 10 mM CaspACE (Promega) in growth media at 37uC, 5% CO 2 for 20 min. After staining, cells were washed twice with PBS and immediately subjected to FACS analysis.

Immunoprecipitation
Immunoprecipitation was carried out as described earlier [32]. Solubilized cells were precleared and incubated with 2 mg of anti-Bax (6A7) or anti-Bak (Ab1) antibody at 4uC for 2 h. Immunoprecipitates were collected by incubating with protein G-Sepharose (Amersham) for 2 h. The pellets were washed intensively with lysis buffer and resuspended in sample buffer before analysis by Western blotting using anti-Bax NT and anti-Bak NT antibodies as described above.

Generation of stable shRNA-expressing HeLa
Stable shRNA-expressing HeLa cells were generated as described before in Kepp et al., 2007. The DNA coding for the shRNA was cloned into pLVTH-M vectors from which it integrated upon viral transfer into the genome of target cells. GFP was used as a marker to select for stable clones. The following sequence was targeted to silence Bim by shRNA expression: TAAGATAACCATTCGTGGG. The efficiency of gene silencing was validated by Western blot analysis and quantitative realtime PCR. The cells transduced with the empty vector were used as controls.

Immunofluorescence microscopy
Cells seeded on coverslips were infected for 15 h, fixed in 3.7% PFA and permeabilized using 0.1% Triton X-100. Nonspecific binding was blocked by using 1% goat serum. The samples were stained using anti-actin (Sigma) and anti-tubulin (molecular probes) followed by detection with fluorochrome-coupled secondary antibodies (Jackson Immuno Research) using a Leica confocal microscope with TCS software or a Zeiss immunofluorescence microscope with ACT software. 3dimensional remodeling was performed using Metamorph and Imaris software.
For apoptosis quantification, fixed cells were stained with 1 mg/ ml Hoechst 33342 (Invitrogen) for 10 min followed by intense washing with PBS. A minimum of 5 fields per slide was analyzed for chromatin condensation using a Zeiss immunofluorescence microscope.

Transmission electron microscopy
Control, infected and infected zVAD treated cells were fixed 15 h post-infection with 2.5% glutaraldehyde, post-fixed with 0.5% osmium tetroxide and contrasted using tannic acid and uranyl acetate. Specimens were dehydrated in a graded ethanol series and embedded in Polybed. Ultrathin sections were analyzed in a Leo 906E transmission electron microscope (Leo GmbH).

Cytoskeletal preparation
To analyze cytoskeleton-associated proteins 1610 6 cells were incubated for 15 min in HBSS (Gibco). All lipidic membranes were destabilized for 5 min at 4uC by incubation with 5 ml high detergent containing Buffer M (1 mM EGTA, 4% PEG 6000, 100 mM PIPES pH 6.9) containing 0.5% Triton X-100. The cytoplasmic and compartmental proteins containing supernatant was removed and the cytoskeleton was washed with cold Buffer M. The remaining proteins were collected in sample buffer and analyzed by Western blotting.

Statistical analysis
The averages and standard errors of the mean as well as the ttests have been calculated using MS Excel. Significance is indicated with ** p , 0,01 and * p , 0,05. If not indicated else wise in the figure legend the data represents at least 3 independent experiments. Video S1 Infection-induced changes in cellular shape. Actin-GFP (Clonetech)-expressing cells were infected and a series of zstacks was acquired every hour using an Olympus spinning disc microscope. The pictures were 3-dimensional remodeled using the Imaris software. Shown is one representative cell undergoing morphological changes.