Yersinia enterocolitica differentially modulates RhoG activity in host cells

Pathogenic bacteria of the genus Yersinia (Y. pestis, Y. enterocolitica and Y. pseudotuberculosis) have evolved numerous virulence factors (termed a stratagem) to manipulate the activity of Rho GTPases. Here, we show that Y. enterocolitica modulates RhoG, an upstream regulator of other Rho GTPases. At the contact site of virulent Y. enterocolitica and host cells, we could visualise spatiotemporally organised activation and deactivation of RhoG. On the one hand, the β1-integrin clustering protein Invasin on the bacterial surface was found to activate RhoG and this promoted cell invasion. On the other hand, active RhoG was downregulated by the type III secretion system effector YopE acting as a GTPase-activating protein (GAP). YopE localised to Golgi and endoplasmic reticulum, and this determined its specificity for RhoG and other selected Rho GTPases. RhoG and its downstream effector module Elmo/Dock180 controlled both Rac1 activation by Invasin and Rac1 deactivation by YopE. We propose that RhoG is a central target of the Yersinia stratagem and a major upstream regulator of Rac1 during different phases of the Yersinia infection cycle.


Introduction
Yersinia pestis, Y. enterocolitica and Y. pseudotuberculosis are widespread bacterial pathogens that elicit plaque, gastroenteritis and enteric lymphadenitis in humans (Naktin and Beavis, 1999;Putzker et al., 2001). After oral ingestion, enteropathogenic Y. enterocolitica and Y. pseudotuberculosis reach distal parts of the small intestine where they invade and cross the mucosa. In the submucosal Peyer-Plaques, the yersiniae proliferate extracellularly and from there they can spread systemically . To elicit and maintain infection, Yersinia has evolved ingenious ways of manipulating host cell functions (Cornelis, 2006;Heesemann et al., 2006;Navarro et al., 2005). Central to the invasive phase of enteropathogenic Yersinia infection is the bacterial surface protein Invasin which triggers outside-in β1-integrin signalling, leading to bacterial permeation of intestinal cells (Wong and Isberg, 2005a). In the subsequent infection phases, yersiniae employ a sophisticated organelle, the type III secretion system (TTSS), to exert antiphagocytic and immunosuppressive effects. The TTSS delivers a collection of effector proteins (Yersinia outer proteins; Yops) into host cells in order to interfere with and reorient various signalling processes Cornelis, 2006;Viboud and Bliska, 2005).
A great variety of bacterial toxins and modulins target Rho GTPbinding proteins, which play pivotal roles in cell regulation (such as organisation of the actin cytoskeleton). The actin cytoskeleton has crucial functions in the infection biology of pathogens, and in cells of innate and adaptive immunity (Aktories and Barbieri, 2005;Ivetic and Ridley, 2004;Vicente-Manzanares and Sanchez-Madrid, 2004). It is well established that the Yersinia stratagem modulates the activity of Rho GTP-binding proteins both for tissue invasion and subversion of the host immune system . The Rho family members RhoA, Rac1 and Cdc42 are activated by Invasin-triggered integrin signalling and mediate cytoskeletal reorganisation, IL-8 release and bacterial invasion (Grassl et al., 2003;Wiedemann et al., 2001;Wong and Isberg, 2005a). By contrast, the cysteine protease YopT inactivates RhoA, Rac1 and Cdc42 by removing their C-terminal isoprenoid moiety (Shao et al., 2002). Furthermore, the serine/threonine kinase YopO/YpkA inhibits the activity of Rac1 and RhoA via a domain that structurally resembles guanine nucleotide dissociation inhibitors (GDIs) for Rho GTPases (Prehna et al., 2006). Finally, YopE mimics a GTPase activating protein (GAP) for Rho GTPases and has been reported to downregulate Rac1, RhoA and Cdc42 in infected cells (Black and Bliska, 2000;Von Pawel-Rammingen et al., 2000). The fact that several Yersinia virulence factors control an overlapping set of Rho GTPases by distinct mechanisms suggests that it is a major goal of the Yersinia stratagem to orchestrate Rho GTPase signalling networks and cascades.
It has been established that downregulation of Rac (Rac1 and presumably Rac2) mediates various immunosuppressive effects of YopE, such as inhibition of phagocytosis, IL-1β production and superoxide anion production Cornelis, 2002;Schotte et al., 2004;Viboud and Bliska, 2005). We have reported previously that the ability of YopE to inhibit Rac regulated actin structures depends on the signal pathway by which Rac is activated (Andor et al., 2001). Recent work shows that the combined action of YopT and YopE produces two distinct pools of Rac1: an active nuclear pool and an inactive cytosolic pool (Wong and Isberg, 2005b). Notably, YopE of some Yersinia serotypes can also be degraded by the host cell proteasome, which causes reactivation of Rac1 (Ruckdeschel et al., 2006). Together, these findings indicate that Rac is the preferred Rho GTPase target of the Yersinia stratagem. This is comprehensible because Rac controls a myriad of cell functions and plays a crucial role in a number of Rho GTPase signalling cascades . One recently recognised way of upstream control of Rac1 is exerted through RhoG and the Elmo/Dock180 module. It has been established that binding of active RhoG to a complex of Elmo and Dock180 stimulates the GEF activity of Dock180, leading to GTP-loading of Rac1 (deBakker et al., 2004;Katoh and Negishi, 2003;Lu and Ravichandran, 2006). Interestingly, RhoG has been implicated in various cell functions known to be subdued by Yersinia, such as integrin-mediated phagocytosis, transendothelial migration and activation of NADPH oxidase (Condliffe et al., 2006;Nakaya et al., 2006;van Buul et al., 2007).
Here, we report that RhoG activity is modulated by Y. enterocolitica. We were able to visualise the spatiotemporal dynamics of RhoG activation and deactivation at the contact site of host cells and virulent Yersinia. The β1-integrin clustering protein Invasin was identified as RhoG activator and the Rho GTPase activating protein YopE as the specific RhoG inhibitor. Modulation of RhoG activity produced parallel changes in Rac1 activity. Thus, our data demonstrate a yin and yang of RhoG modulation by the Yersinia stratagem that may be adapted to the different phases of the Yersinia infection cycle in vivo.

Differential modulation of RhoG activity by Yersinia enterocolitica
In order to investigate whether RhoG is regulated by Y. enterocolitica, we established a biosensor for monitoring active GTP-bound RhoG in cells. It has been reported that the armadillo repeats at the N-terminus of Elmo (Elmo2NT) constitute a binding domain for GTP-bound RhoG but not for other Rho GTPases (deBakker et al., 2004;Katoh and Negishi, 2003). We confirmed by pulldown that GST-Elmo2NT binds to constitutively active RhoGV12 but not to other constitutively active Rho GTPases or to dominant-negative RhoGN17 (Fig. l). Upon co-expression with RhoGV12, GFP-Elmo2NT redistributed to the perinuclear area of human umbilical vein endothelial cells (HUVEC) where it colocalised with RhoGV12 (Fig. 2). By contrast, co-expression with RhoGN17 did not alter the normally diffuse cellular distribution of GFP-Elmo2NT (Fig. 2). Moreover, the RhoG guanine nucleotide exchange factors (GEFs) TrioD1 and SH3-containing guanine nucleotide exchange factor (SGEF) caused translocation of GFP-RhoG and GFP-Elmo2NT to dorsal ruffles in HUVEC ( Fig. 2; see Fig. S1 in the supplementary material). TrioD1 and SGEF are RhoGspecific GEFs that have been shown to induce GTP-loading and activation of RhoG in vitro and in cells (Blangy et al., 2000;Ellerbroek et al., 2004). Rac1V12 and the Rac-specific GEF Tiam1 did not induce GFP-Elmo2NT redistribution in HUVEC (B.R., A.R., M.H., K.R. and M.A., unpublished). Together, these experiments indicate that GFP-Elmo2NT faithfully monitors the localisation of active RhoG in cells.
To visualise the spatiotemporal dynamics of RhoG activation, we performed live imaging of Yersinia-infected HUVEC expressing fluorescent RhoG and/or Elmo2NT. The recordings indicate that GFP-RhoG rapidly and continuously encloses single bacterial rods of Yersinia WA-C. Formation of a full circumferential phagocytic cup took about 1-2 minutes (mean 70±54 seconds, n=3) (single frames of a representative movie are presented in Fig. 3B). Endogenous RhoG also translocated to the Yersinia WA-C cell contact site (supplementary material Fig. S2A). In sharp contrast to the fast and continuous recruitment of GFP-RhoG by the WA-C bacteria, GFP-RhoG appeared to alternatively advance and recede along the rods of Yersinia WA-314. This back and forth process could take up to 23 minutes (14.6±8.5 minutes, n=3) (single frames of a representative movie are shown in Fig. 3B) (see also supplementary material Movie 1), although in all cases recorded it eventually led to formation of a closed phagocytic cup. Such a disordered and protracted phagocytic cup assembly could also be seen with GFP-Elmo2NT (Fig. 3B). These data therefore visualise subversion of RhoG activation at the bacteria cell contact site. Further experiments revealed that mCherry-RhoG and GFP-Elmo2NT translocated in parallel to Yersinia WA-C. Whereas GFP-Elmo2NT stayed at the phagocytic cups only until phagocytic cup closure, GFP-RhoG remained there for extended time periods (B.R., A.R., M.H., K.R. and M.A., unpublished). We conclude that RhoG is recruited by the bacteria in its activated form, is deactivated shortly after phagocytic cup closure but then remains at the nascent phagosome.
To further verify regulation of RhoG activity by Y. enterocolitica, we performed GST-Elmo2NT pull-down assays. As demonstrated in Fig. 4A, infection of GFP-RhoG transfected Cos-7 cells with Yersinia WA-C, but not Yersinia WA-314, produced an increased level of GTP-bound GFP-RhoG. Activation of GFP-RhoG was also detected in Cos-7 cells treated with GST-Invasin-397 beads or Invasin-E. coli but not with YadA-E. coli (Fig. 4A). YadA is a matrix binding adhesin of enteropathogenic yersiniae (Roggenkamp et al., 1995). In HeLa cells, which express relatively high levels of endogenous RhoG, we could show activation of endogenous RhoG by Yersinia WA-C but not WA-314. Endogenous RhoG was also activated by Invasin-E. coli in these cells (Fig. 4B). Together these experiments indicate that in Yersinia infected cells RhoG is activated by Invasin-induced signalling but then downregulated by a TTSStranslocated Yop effector.
YopE has not formally been shown to operate as a GAP for RhoG, although our results are consistent with this notion. To demonstrate that YopE stimulates the intrinsic GTPase activity of RhoG, we performed an in vitro GTP-hydrolysis assay. The data presented in Fig. 5D clearly demonstrate that YopE is a bona fide GAP for RhoG and acts as efficiently on RhoG as on Rac1 in vitro.
Having identified YopE as a potent deactivator of RhoG, we next tested whether YopE is sufficient to inhibit RhoG recruitment by Yersinia. HUVEC-expressing EGFP-RhoG were infected with WA-C(pTTSS+pYopE) or WA-C(pTTSS+pYopER144A), and GFP-RhoG recruitment to phagocytic cups was evaluated. A timedependent 30 to 50% reduction of RhoG recruitment by the Yersinia strain translocating YopE when compared with the strain translocating YopER144A was detected (Fig. 6). Consistently, recruitment of GFP-Elmo2NT was reduced by more than 50% by the YopE-translocating versus the YopER144A-translocating Yersinia strain (supplementary material Fig. S5). In summary, we conclude that YopE is a specific and highly effective inihibitor of RhoG activity during Yersinia infection.
Previously, a region comprising amino acid residues 54-75 (termed membrane localisation domain; MLD) was found to mediate intracellular membrane binding of YopE (Krall et al., 2004). Consistently, a YopE construct lacking the MLD (GFP-YopER144AΔMLD) displayed a cytoplasm-like fluorescence and no signs of Golgi/ER localisation (supplementary material Fig.  S3C). Recombinant YopEΔMLD showed the same specificity and activity as native YopE in the in vitro GAP assay (Fig. 5D). In order to clarify the function of YopE compartmentalisation, we tested the effect of WA-C(pTTSS+pYopEΔMLD) on TrioD1-activated RhoG in Cos-7 cells. As demonstrated in Fig. 5C, right and left panels, YopEΔMLD and YopE were equally effective in downregulating RhoG. Furthermore, Cdc42 was also equally well deactivated by YopE and YopEΔMLD (supplementary material Fig. S4A). However, whereas YopEΔMLD abrogated baseline RhoA activity in Cos-7 cells YopE did not have an effect on RhoA (Fig. 8). Similarly, YopEΔMLD could effectively deactivate TC10, whereas YopE had only a minor effect on TC10 (supplementary material Fig. S4B). Together, these experiments suggest that the substrate range and activity of YopE is crucially determined by its compartmentalisation in the Golgi/ER.

RhoG controls activation of Rac1 by Invasin and deactivation of Rac1 by YopE
The question arises of how does modulation of RhoG fit in the virulence strategy of Yersinia? It is noticeable that RhoG can control Rac1 via the Elmo/Dock180 module and that Rac1 is the preferred Rho GTPase target of Yersinia Katoh and Negishi, 2003). We thus asked whether RhoG mediates the effects of the Yersinia stratagem on Rac1. To answer this question, Cos-7 cells in which RhoG was knocked down by RNAi were infected with WA-C or Invasin-E. coli and Rac1 activity was determined by GST-PAK-CRIB pulldown. Knock down of RhoG not only inhibited baseline Rac1 activity but also blocked Invasinstimulated Rac1 activation almost quantitatively (Fig. 9A,B). Inhibition of baseline and Invasin-stimulated Rac activation could also be obtained by expression of dominant-negative Elmo-T618 in Cos-7 cells (B.R., A.R., M.H., K.R. and M.A., unpublished) (Katoh and Negishi, 2003). The finding that cell invasion of Yersinia WA-C was reduced by RhoG RNAi proves the functional significance of the RhoG signal pathway (Fig. 9C).
We next investigated to what extent YopE-induced Rac1 inhibition is in fact mediated by downregulation of RhoG and the Elmo/Dock180 module. Infection of Cos-7 cells with WA-C(pTTSS+pYopE) largely reduced the baseline activation level of Rac1 (Fig. 9D). As already found for knock down of RhoG by RNAi (Fig. 9A), expression of Elmo-T618 reduced the baseline level of activated Rac1 in Cos-7 cells (Fig. 9D). Thus, YopE, RhoG RNAi and dominant-negative Elmo-T618 induce a similar degree of Rac1 inhibition. Notably, infection of the Elmo-T618-expressing cells with WA-C(pTTSS+pYopE) did not produce a further reduction of Rac1 activity (Fig. 9D). We propose that under our experimental conditions YopE downregulates baseline Rac1 activity for the most part through inhibition of RhoG. By comparison, RhoA activity was neither affected by YopE nor by Elmo-T618 (B.R., A.R., M.H., K.R. and M.A., unpublished).
To clarify whether or not YopE can directly inactivate Rac1, we induced Rac1 activation in HUVEC by expression of the Racspecific exchange factor Tiam1. Infection with WA-C(pTTSS+pYopE) caused inhibition but not abrogation of Tiaminduced Rac1 activity (Fig. 9E). These data clearly show that YopE  can directly inhibit Rac1 activity by competition with Rac exchange factors.
We conclude that YopE inactivates Rac1 by direct and indirect mechanisms. Clearly, YopE can compete with Rac GEFs and can directly stimulate GTP-hydrolysis in Rac1. In addition, YopE can induce GTP-hydrolysis in RhoG and thereby downregulate Elmo/Dock180 signalling to Rac1, a mechanism that appears to play a prominent role in Yersinia infection.
Together these experiments suggest that by regulating RhoG Yersinia can exert an additional way of control over the Rac GTPase.

Journal of Cell Science 122 (5)
This most certainly will affect multiple host cell processes during Yersinia infection.

Discussion
We report here that Y. enterocolitica differentially regulates the activation level of the Rho GTPase RhoG. RhoG becomes activated by Invasin-triggered β1-integrin signalling and this mediates bacterial invasion. RhoG has previously been implicated in phagocytosis and bacterial invasion. A signalling pathway involving RhoG, its GEF Trio and the Rac GEF-complex Elmo/Dock180 was shown to regulate phagocytosis of apoptotic cells (deBakker et al., 2004). The TTSS effector SopB of Salmonella enterica was reported to activate RhoG via SGEF and thereby stimulate actin remodelling (Patel and Galan, 2006). Furthermore, the Shigella effector IpgB1 mimics RhoG to activate Elmo/Dock180, which induces membrane ruffling and bacterial invasion (Handa et al., 2007). Similarly, our data suggest that Invasin-triggered integrin signalling causes activation of the RhoG/Elmo/Dock 180 module. Hence, RhoG evolves as a central target of bacterial virulence factors promoting host cell invasion. Clearly, besides the RhoG/Rac1 signalling system, RhoA-and Cdc42-regulated pathways can also contribute to bacterial invasion Patel and Galan, 2006). RhoG may exclusively act through its close relative Rac1 for mediating bacteria-induced cell responses. However, RhoG may also have an individual function in bacterial invasion given the reported Rac1-independent RhoG effects on the cytoskeleton (Meller et al., 2008) and on intracellular vesicle transport (Prieto-Sanchez et al., 2006;Tanaka et al., 2007).
For visualising the spatiotemporal dynamics of RhoG activation at the Yersinia cell contact site, we employed spinning disc microscopy of GFP-Elmo2NT. Several pieces of evidence indicate that localisation of intracellular GFP-Elmo2NT properly monitors active RhoG. Elmo2NT binds to active RhoG but not to other Rho GTPases. Constitutively active, but not dominant-negative, RhoG induced GFP-Elmo2NT redistribution to Golgi and ER, and two RhoG-GEFs stimulated its redistribution to dorsal ruffles. GFP-Elmo2NT and mCherry-RhoG were co-recruited to phagocytic cups. The finding that, shortly after phagocytic cup closure, GFP-Elmo2NT vanished whereas mCherry-RhoG remained at the phagosome, supports the notion that Elmo2NT associates with an active subfraction of RhoG that functions in the early phases of phagocytosis/invasion (B.R., A.R., M.H., K.R. and M.A., unpublished).
Live cell imaging revealed that virulent Y. enterocolitica considerably protracts phagocytic cup assembly and causes an artificial alternation of RhoG activation and deactivation. We could identify YopE as the sole RhoG inhibitor of virulent yersiniae. YopE reportedly has multiple effects on host cell regulation, such as inhibition of phagocytosis, IL-1β production and superoxide anion production Cornelis, 2002;Schotte et al., 2004;Viboud and Bliska, 2005), and it also affects Yop translocation by regulating the TTSS pore (Mejia et al., 2008). Although most YopE effects have been attributed to inhibition of Rac1 Cornelis, 2002;Schotte et al., 2004;Viboud and Bliska, 2005) and we in fact confirmed that YopE can directly inhibit Rac1, we demonstrate here that YopE abrogates baseline-and Invasin-stimulated Rac1 activity by inhibition of RhoG. Hence, there are at least two mechanisms by which YopE can modulate Rac1 activity in host cells: by direct downregulation of Rac1 and by blocking RhoG and the ensuing Elmo/Dock180 signalling to Rac1. Together with the previous findings that: (1) Fig. 9. Invasin stimulates and YopE inhibits Rac1 activity via RhoG. (A) Cos-7 cells were transfected with siRNA for RhoG and at indicated time points equal amounts of protein were analysed by western blotting using anti-RhoG and anti-Rac1 antibody. RhoG siRNA effectively reduced the level of endogenous RhoG protein at 48 hours but had no effect on Rac1 protein level (left panel). Cos-7 cells were transfected with non-targeting (nt) siRNA or with RhoG siRNA for 48 hours and then subjected to GST-PAK-CRIB pull-down assay. Rac1 was detected by western blotting using anti-Rac1 antibody (right panel). (B) HUVEC were left untransfected or transfected with RhoG siRNA (indicated by stars) for 48 hours and then infected or not with WA-C or Invasin-E. coli for 15 minutes. The level of active Rac1 was determined by GST-PAK-CRIB pull-down and western blotting using anti-Rac1 antibody. (C) HUVEC were transfected with RhoG siRNA or non-targeting (nt) siRNA for 48 hours and then infected with Yersinia WA-C for 60 minutes. Invasion of bacteria was determined using a double fluorescence method (see Materials and Methods). Bars represent mean±s.d. (error bars) of three independent experiments with at least 60 cells analysed per experiment. (D) Cos-7 cells were transfected with myc-Rac1 (lane 1 and 2) or co-transfected with myc-Rac1 and HA-Elmo-T618 (lane 3 and 4) and then infected with WA-C(pTTSS+pYopE) for 2 hours. The level of active myc-Rac1 was determined by GST-PAK-CRIB pull-down and western blotting using anti-myc antibody. HA-Elmo-T618 was detected with anti-HA antibody. (E) HUVEC were transfected with plasmids expressing myc-Rac1 (lane 1 and 2) or myc-Rac1 and HA-Tiam1 (lane 3 and 4), and infected with Yersinia strain WA-C(pTTSS+pYopE) for 2 hours. The level of active Rac1 was determined by GST-PAK-CRIB pull-down as described in D. some signalling pathways to Rac are not blocked by YopE; (2) cooperation of Yersinia YopE and YopT can lead to differentially activated and localised pools of Rac in cells; and (3) YopE of some Yersinia serotypes is degraded in host cells leading to reactivation of Rac1, a highly sophisticated strategy evolves by which Yersinia manipulates the Rac1 signalling network (Andor et al., 2001;Hentschke et al., 2007;Ruckdeschel et al., 2006;Wong and Isberg, 2005b).
We also identified a function for the Golgi and ER localisation of YopE. YopE lacking its membrane localisation domain (YopEΔMLD) was at least as active as the wild-type YopE in inhibiting RhoG and Cdc42, and newly acquired the capability to inhibit RhoA. In this regard, YopE behaves differently from its homolog ExoS from Pseudomonas aeruginosa, for which the substrate range is restricted upon removing the MLD (Zhang et al., 2007). Interestingly, phospholipids can switch the substrate specificity of GAPs (Ligeti et al., 2004). Thus, the different lipid environments of membrane-bound YopE versus cytoplasmic YopEΔMLD may also influence their Rho GTPase specificities.
In summary, we propose that during the invasive phase of Yersinia infection, when Yops are not yet expressed, Invasin-triggered RhoG activation mediates bacterial internalisation, and this occurs at least partly by stimulating Rac1 via the Elmo/Dock180 module. During the subsequent antiphagocytic and immunosuppressive phase of the infection, RhoG also becomes activated by Yersinia but this may be triggered by a phagocytosis program of cells of innate immunity. At this point, RhoG and Rac1 are downregulated by YopE in a compartmentalised and differential fashion (Fig. 10). Rac1, which has pleiotropic functions in the biology of Yersinia infection and in immune cells, is accessorily modulated by YopT and YopO. It is likely that YopE, YopT and YopO synergise in producing differentially active and compartmentalised subsets of Rac1 and potentially of other Rho GTPases in host cells. Using this strategy, Yersinia may allow Rho GTPase activation in selective cell compartments and signalling pathways, and in parallel downregulate Rho GTPases in other pathways. For example, by leaving open the bacterial translocation pore, Rac1 activity may be admitted whereas, in parallel, Rac1 pathways that regulate immune cell responses are inhibited. The relevance of such strategies for the in vivo infection process of Yersinia is speculative at the moment but needs to be investigated in future studies.

Bacterial strains and cell infection
For cell infection with Y. enterocolitica strains (Table 1), overnight bacterial cultures grown at 27°C in Luriani-Bertani broth (LB) containing the appropriate antibiotics were diluted 1:20 in fresh LB and grown for additional 90 minutes at 37°C to allow for expression of the type III secretion machinery and Yops. For infection with E. coli HB101 inv+ and DH5α-yadA+, bacteria were grown overnight at 37°C in LB containing the appropriate antibiotics, diluted 1:20 in fresh LB and grown for additional 2 hours at 37°C. To synchronize infection, bacteria were attached to cells by centrifugation at 200 g for 2 minutes at a ratio of 10 to 100 bacteria per cell. Latex beads (1 μm diameter, sulphate microspheres, Invitrogen, Karlsruhe, Germany) were coated with GST or GST-Inv397 as described  and centrifuged onto the cells at 200 g for 3 minutes at a bead to cell ratio of 100:1. At designated time points, cells were washed twice in ice-cold PBS and processed for immunofluorescence or pull-down assays. For live-cell imaging, bacteria were pipetted into the medium and allowed to sediment onto cells. Fig. 10. Model for modulation of RhoG activity by Yersinia. During the invasive phase of infection, Yersinia binds to integrins via Invasin, which triggers activation of RhoG and stimulation of Rac1 through the Elmo/Dock180 module. Activation of RhoG and Rac1 leads to cellular responses, such as invasion, phagocytosis, caspase 1 activation and production of interleukins. During the subsequent immunosuppressive phase, Yersinia injects effector Yops through its TTSS. The effector YopE acting as a GAP inactivates RhoG and Rac1, and reverses the specific cellular responses. France). The gene encoding YopE was amplified by PCR from Y. enterocolitica serotype O9 strain E40 (Hentschke et al., 2007) and cloned into vectors pEGFP-N1 (Clontech, St.-Germain-en-Laye, France), pGEX4T-2 (GE Healthcare, Munich, Germany) and pmCherry (gift of Dr Marek Drab, Max-Planck Institute for Infection Biology, Berlin, Germany). YopER144A was generated using QuikChangeII sitedirected mutagenesis kit (Stratagene, Amsterdam, Netherlands) and YopEΔMLD using Phusion site directed Mutagenesis kit (Finnzymes, Espoo, Finland). The genes encoding YopER144A and YopEΔMLD together with SycE were subcloned into vector pACYC184 (N.E. Biolabs, Frankfurt, Germany) and transformed into Y. enterocolitica strain WA-C(pTTSS), resulting in strains WA-C(pTTSS+pYopER144A) and WA-C(pTTSS+pYopEΔMLD), respectively (Table 1). All GST-tagged proteins were purified from BL21 E. coli using glutathione-Sepharose 4B (GE Healthcare, Munich, Germany) as described elsewhere (Trasak et al., 2006). Where necessary, proteins were dialysed against PBS overnight at 4°C.

In vitro GAP assay
In vitro GAP assay was performed with the RhoGAP assay Biochem kit (Cytoskeleton, Denver, USA). RhoG, YopE and YopEΔMLD were expressed and purified as GSTfused proteins, recombinant Rac1, Ras and p50RhoGAP were part of the kit. Proteins were incubated at 37°C with 200 mM GTP. After 20 minutes the reaction was stopped and green colour development of free phosphate was measured at 650 nm.

Immunofluorescence and live cell imaging
Cells were fixed in 3.7% formaldehyde in PBS for 10 minutes and permeabilized with ice-cold acetone for 5 minutes. Actin was stained with Alexa 568-labeled phalloidin (Molecular Probes, Karlsruhe, Germany), Golgi apparatus with anti-GS27 (BD Biosciences, Heidelberg, Germany) and anti p230 (BD-Biosciences, Heidelberg, Germany), endoplasmatic reticulum with anti-calreticulin (Stressgen, Vancouver, Canada), myc-tagged constructs with anti-myc (Cell Signaling, Frankfurt, Germany) and GFP-tagged constructs with anti-GFP (BD-Biosciences, Heidelberg, Germany). The anti-RhoG antibody was a generous gift from Dr Martin Schwartz (University of Virginia, Charlottesville, VA). Specificity of endogenous RhoG immunostaining using this antibody was verified in siRNA RhoG knock down cells (supplementary material Fig. S2B).
YopE was stained with a rabbit polyclonal antibody (Ruckdeschel et al., 2006). Secondary antibodies were Alexa-488-or Alexa-568-labeled goat anti-mouse or goat anti-rabbit IgG (Molecular Probes, Karlsruhe, Germany). Coverslips were mounted in MOWIOL (Calbiochem, Darmstadt, Germany) containing 0.18% pphenylendiamine (Sigma-Aldrich, Munich, Germany) as anti-fading reagent and sealed with clear nail polish. Images of fixed samples were acquired with a confocal laser-scanning microscope (Leica DM IRE2 with a Leica TCS SP2 AOBS confocal point scanner) equipped with an oil-immersion plan Apo 63ϫ NA 1.4 objective. Acquisition and processing of images was performed with Leica Confocal Software (Leica, Wetzlar, Germany). For the internalization assay, bacteria were quantified with a double fluorescence staining method as described (Schröder et al., 2006). For confocal time-lapse microscopy, HUVEC were transfected and seeded onto glassbottomed dishes (MatTek, Ashland, USA) 1 day before the experiment. For live imaging, cells were kept in a microscope chamber at 37°C and 5% CO 2 . For assessing the kinetics of phagocytic cup formation, t=0 was defined as the first visible accumulation of a GFP-protein at the bacterium and phagocytic cup closure is defined as the time point when a completely closed cup can be seen for the first time. Images were acquired with a spinning disc confocal system (Spinning disc CSU22, Yokogawa, Japan) fitted on a Zeiss Axiovert 200M microscope with a temperature-and CO 2controllable environmental chamber (Solent Scientific, Regensworth, UK), oil immersion plan Apo 63ϫ NA 1.4 objective and a CCD camera (EM-CCD C-9100-2, Hamamatsu, Japan). Acquisition and processing of images was performed with Volocity Software (Improvision, Coventry, UK).

siRNA knock down
A pool of four siRNAs against RhoG and a pool of siControl nontargeting siRNAs were obtained from the Dharmacon siRNA collection (Lafayette, USA). HUVEC were transfected with the Amaxa Nucleofector system. Cos-7 were transfected with X-treme Gene siRNA transfection reagent (Roche, Mannheim, Germany).