VipD of Legionella pneumophila Targets Activated Rab5 and Rab22 to Interfere with Endosomal Trafficking in Macrophages

Upon phagocytosis, Legionella pneumophila translocates numerous effector proteins into host cells to perturb cellular metabolism and immunity, ultimately establishing intracellular survival and growth. VipD of L. pneumophila belongs to a family of bacterial effectors that contain the N-terminal lipase domain and the C-terminal domain with an unknown function. We report the crystal structure of VipD and show that its C-terminal domain robustly interferes with endosomal trafficking through tight and selective interactions with Rab5 and Rab22. This domain, which is not significantly similar to any known protein structure, potently interacts with the GTP-bound active form of the two Rabs by recognizing a hydrophobic triad conserved in Rabs. These interactions prevent Rab5 and Rab22 from binding to downstream effectors Rabaptin-5, Rabenosyn-5 and EEA1, consequently blocking endosomal trafficking and subsequent lysosomal degradation of endocytic materials in macrophage cells. Together, this work reveals endosomal trafficking as a target of L. pneumophila and delineates the underlying molecular mechanism.


Introduction
Legionella pneumophila is an opportunistic human pathogen that replicates inside macrophages, which are at the front line of immune defense. This Gram-negative bacterium causes Legionnaires' disease characterized by severe pneumonia or less acute Pontiac fever. By phagocytosis, the bacteria are enclosed in a membrane-bound vacuole, called Legionella-containing vesicle (LCV). This vesicle evades the endocytic pathway to avoid fusion with lysosomes [1], and becomes the growth and replication niche for the bacteria [2,3]. The intracellular survival and replication depend on the Dot/Icm type IV secretion system of the bacterium, which translocates about 270 effector proteins into the host cytosol [4,5]. Understanding of detailed molecular mechanisms of the L. pneumophila effectors has been achieved for a number of proteins, including SidM (substrate of Icm/Dot transporter M; also known as DrrA) [6][7][8][9][10][11][12], LpGT (L. pneumophila glucosyltransferase; also known as Lgt1) [13][14][15], AnkX (Ankyrin repeat protein X) [16][17][18] and others as reviewed recently [19].
VipD (vacuolar protein sorting inhibitor protein D) is one of the L. pneumophila effector proteins, which interrupts Golgi-to-vacuole trafficking of three yeast proteins (carboxypeptidase S, carboxypeptidase Y and alkaline phosphatase) as well as endoplasmic reticulum (ER)-to-Golgi trafficking of carboxypeptidase Y when expressed in Saccharomyces cerevisiae [20]. VipD contains an N-terminal lipase domain which shares sequence homology with patatin, a phospholipase in potato tuber having phospholipase A and lysophospholipase A activities [21]. A similar lipase domain is present in two other L. pneumophila effector proteins VpdA and VpdB [22] and in ExoU of Pseudomonas aeruginosa, which is a potent secreted cytotoxin [23,24]. On the other hand, their C-terminal domains do not exhibit sequence homology with each other or with any functionally annotated protein domain. Previously, overexpression of VipD was shown to be mildly toxic to 293T cells and S. cerevisiae, and its toxicity was only partially dependent on the putative lipase activity of the protein [22]. Moreover, a VipD fragment lacking the N-terminal lipase domain interfered with vesicle transport in S. cerevisiae to a much greater extent than full-length VipD did [20], indicating that the C-terminal domain is critical for the function of VipD. But, how the C-terminal domain of VipD perturbs vesicle trafficking in yeast is unknown. It is also unknown whether VipD may manipulate intracellular trafficking in macrophages, the major mammalian host cells of L. pneumophila.
We undertook an integrative approach involving X-ray crystallography, biochemistry and cellular imaging to understand whether and how VipD might affect mammalian host cells. We show that the C-terminal domain of VipD tightly binds to the GTP-bound form of Rab5 and Rab22, blocks their interactions with three downstream effector molecules, and inhibits endocytic trafficking in mouse macrophages. Together, this study demonstrates that VipD targets and interferes with endosomal membrane trafficking in mammalian host cells.

VipD adopts a two-domain fold
Full-length VipD was crystallizable, but the X-ray diffraction of the crystals was too poor for structure determination. Various attempts were made to improve the crystal quality, and the successful trial was to employ a truncated VipD lacking C-terminal 46 residues and to dehydrate resulting crystals with 30% glycerol at 220uC. The structure of this truncated version of VipD, referred to as VipD , was determined at 2.9 Å resolution (Table 1). VipD(1-575) folds into two domains which are roughly discernable: the N-and C-terminal domains, designated as VipD(1-316) and VipD(316-575), respectively ( Figure 1A). Ala316 is at the boundary of the two domains and located in the middle of the structure lengthwise ( Figure 1A). The two domains interact with each other mostly through secondary structural elements. b1 and b2 of VipD(1-316) form a ''mini'' bsheet together with b11 of VipD(316-575) in the C-terminal domain. Likewise, b10 of VipD(316-575) is a part of the central bsheet in the N-terminal domain. In addition, a14 of VipD(316-575) interacts with a2 of VipD(1-316) ( Figure 1A). These observations suggested that division of VipD into the two fragments containing residues 1-316 or 316-575 would result in misfolded proteins. However, both VipD(1-316) and VipD(316-575) or VipD(316-621) produced in E. coli were soluble and purifiable.
A search for similar structures in the Protein Data Bank with the program Dali [25] showed that the N-terminal domain is most homologous to patatin (PDB entry: 1OXW) and cytosolic phospholipase A 2 (cPLA 2 ; PDB entry: 1CJY) with the Z-scores of 14.0 and 11.4, respectively ( Figure S1A). In particular, the two residues of cPLA 2 (Ser228 and Asp549), which form the catalytic dyad [26], are closely superposable on Ser73 and Asp288 in VipD ( Figure 1B). Moreover, the Gly196-Gly-Gly-Phe-Arg200 sequence, which forms the oxyanion hole in cPLA 2 , is also present in VipD as a Gly42-Gly-Gly-Ala-Lys46 sequence at spatially the same location ( Figure 1B). In cPLA 2 , the active site groove containing the catalytic dyad is partially covered by loop aH-aI. Likewise, a similar groove covered by loop b10-a14 is present in VipD ( Figure S2). These features indicate that VipD is a catalytically active phospholipase A 2 . However, whether VipD has an intrinsic phospholipase A 2 activity or not has been unsettled [22,27]. We examined a phospholipase A 2 activity of VipD by using an artificial fluorogenic phospholipid substrate red/green BODIPY PC-A2 (specific for PLA 2 enzyme), and show here that VipD has a phospholipase A 2 activity ( Figure 1C). Alanine substitution of Ser73 or Asp288 abrogated the lipase activity of VipD, demonstrating that the two residues indeed form a catalytic dyad ( Figure 1C).
VipD(316-575) contains ten a-helices and two short b-strands. This domain is not obviously homologous to any of the known protein structures in the Protein Data Bank. The best match (Zscore: 4.6) in the Dali search was the structure of the Vps9 domain of Rabex-5 (PDB entry: 1TXU), which is a guanine nucleotide exchange factor (GEF) for Rab5, Rab21 and Rab22 [28,29]. Superposition of the two structures showed only a gross similarity in the spatial arrangement of five out of ten a-helices in VipD(316-575) ( Figure S1B), providing only an unconvincing clue for the function of the C-terminal domain.

VipD localizes to early endosomes via the C-terminal domain
A clue for the biochemical function of the C-terminal domain of VipD was obtained by investigating the subcellular localization of VipD. In HeLa cells, full-length VipD, VipD  or VipD(316-621) was transiently expressed, each as a fusion protein containing yellow fluorescent protein (YFP) at the C-terminus. Full-length VipD and VipD(316-621) exhibited a similar fluorescence pattern, which was indicative of endosomal localization ( Figure 2A). To elaborate this observation further, full-length VipD or VipD(316-621) was coexpressed with the early endosomal markers Rab5b and Rab22a and also with the ER-to-Golgi trafficking regulator Rab1a [30], respectively, in HeLa cells and in RAW264.7 macrophages. The GTPase-defective constitutively active forms, Rab5b(Q79L), Rab22a(Q64L) and Rab1a(Q70L), were employed, all tagged with cyan fluorescent protein (CFP). Both full-length VipD and VipD(316-621) colocalized with Rab5b(Q79L) and Rab22a(Q64L), but not with Rab1a(Q70L), in both types of cells ( Figures 2B and S3). In contrast, VipD(1-316) was evenly dispersed throughout cells with a noticeable enrichment at the plasma membranes ( Figure 2A). Notably, the characteristic tubular structures of endosomes observed with the expression of Rab22a(Q64L) alone ( Figure S4A) [31] disappeared when this Rab protein was coexpressed together with full-length VipD or VipD(316-621), while their formation was unaffected by the expression of VipD(1-316) ( Figure S4B). We additionally noted that Rab22a(Q64L) colocalized with Rab5b(Q79L) without inducing the tubular structures when the two proteins were coexpressed in both types of cells (Not shown). We also found that VipD colocalized with the wild-type forms of Rab5b and Rab22a

Author Summary
Legionella pneumophila is a pathogen bacterium that causes Legionnaires' disease accompanied by severe pneumonia. Surprisingly, this pathogen invades and replicates inside macrophages, whose major function is to detect and destroy invading microorganisms. How L. pneumophila can be ''immune'' to this primary immune cell has been a focus of intensive research. Upon being engulfed by a macrophage cell, L. pneumophila translocates hundreds of bacterial proteins into this host cell. These proteins, called bacterial effectors, are thought to manipulate normal host cellular processes. However, which host molecules and how they are targeted by the bacterial effectors are largely unknown. In this study, we determined the three-dimensional structure of L. pneumophila effector protein VipD, whose function in macrophage was unknown. Ensuing analyses revealed that VipD selectively and tightly binds two host signaling proteins Rab5 and Rab22, which are key regulators of early endosomal vesicle trafficking. These interactions prevent the activated form of Rab5 and Rab22 from binding their downstream signaling proteins, resulting in the blockade of endosomal trafficking in macrophages. The presented work shows that L. pneumophila targets endosomal Rab proteins and delineates the underlying molecular mechanism, providing a new insight into the pathogen's strategies to dysregulate normal intracellular processes.
( Figure S5A), which cycle between the endosomal membrane and the cytosol [30]. Finally, like Rab5b(Q79L), full-length VipD did not localize to lysosomes, as probed by the lysosomal marker Lysotracker Red ( Figure S5B). These results convincingly indicat-ed that VipD localizes to early endosomes via the C-terminal domain of the protein. In addition, the precise overlaps of the two different fluorescence images suggested that the C-terminal domain of VipD may directly interact with the two Rabs.

VipD tightly binds to specific endosomal Rabs
The possibility of the direct interactions of VipD with Rab5b and Rab22a was probed using Rab5b(1-190;Q79L) and Rab22a(1-175;Q64L), and a wild-type version of the two Rabs. These Rab proteins were C-terminally fused to a (His) 10 -tagged cysteine protease domain (CPD) to improve the solubility of the target proteins [32]. Indeed, VipD interacted with the GTP-bound Q-to-L mutant form of the two Rabs in a (His) 10 Figure 3D). The (His) 10 pull-down assay was also performed with Rab5a and Rab5c, the two other isoforms of Rab5. VipD bound to the two forms of Rab5a and Rab5c similarly as it did to Rab5b: tightly to the active form and weakly to the inactive form ( Figure 3A; first and third panels). We also examined the interaction between VipD and Rab22b (also known as Rab31), the other isoform of Rab22. VipD tightly bound to both the GTPbound and the GDP-bound forms of Rab22b(1-175), as it did to the two forms of Rab22a(1-175) ( Figure 3B; second panel). We, in turn, examined whether these isoforms of Rab5 and Rab22 colocalize with VipD in HeLa cells. The active forms of Rab5a and Rab5c indeed colocalized with VipD, as Rab5b(Q79L) did ( Figure S5C). However, Rab22b(Q64L) did not colocalize with VipD ( Figure S5D). Rab22b is known to be largely associated with the trans-Golgi network in HeLa cells [33]. Consistently, the subcellular distribution of Rab22b(Q64L) overlapped only partially with that of the endosomal marker Rab22a(Q64L) ( Figure  S5D). These observations implied that the endosomal localization of VipD does not simply depend on the interaction with Rab proteins. To test this notion, a set of small interfering RNAs (siRNAs) was prepared which blocked the expression of Rab5a, Rab5b, Rab5c and Rab22a [31,[34][35][36]. Treatment of HeLa cells with each siRNA alone or together did not block the endosomal localization of VipD ( Figure S6), indicating that VipD localizes to endosomes through an as yet unknown mechanism and then interacts with the endosomal Rab proteins.

VipD is a very weak GEF antagonist and not a GAP
Given the protein-binding analyses and the minor structural similarity between the C-terminal domain of VipD and the Vps9 domain of Rabex-5 ( Figure S1B), we suspected that VipD might have a GEF activity toward Rab5 and Rab22. This possibility was tested by fluorescence resonance energy transfer assay using 29/39-O-(N9-methylanthraniloyl)-GDP (mant-GDP)-loaded Rab5b  and Rab22a(1-175). VipD exhibited no observable GEF activity toward the two Rabs (Not shown; indirectly shown in Figure 4A), and thus it is not a GEF for the two Rabs. An alternative possibility that VipD might competitively inhibit a cellular GEF for Rab5 and Rab22 was examined by performing a GEF activity assay with Rab5b(1-190):mant-GDP, Rab22a(1-175):mant-GDP and the Vps9 domain (residues 132-397) of Rabex-5, which is a strong and a comparatively weak GEF for Rab5 and Rab22, respectively [28]. VipD noticeably but very weakly inhibited the GEF activity of the Vps9 domain; 1000 molar excess of VipD over the Vps9 domain decreased k cat /K M by only about two folds for Rab5b   (Figure 4A; left panel). VipD inhibited the weak GEF activity of Rabex-5 toward Rab22a more evidently. However, also in this case, only five folds decrease of k cat /K M was detected when VipD was present at 1000 molar excess over Rabex-5 ( Figure 4A; right panel). These results indicate that VipD can interfere with the GEF activity of Rabex-5 but only slightly. These inhibitory effects presumably arise from the binding affinity of VipD for the GDP-bound forms of the two Rabs ( Figure 3). Whether VipD could function as a GTPaseactivating protein (GAP) for Rab5b was also tested by employing the GAP domain of RabGAP-5 (residues 1-451), a specific cellular GAP for Rab5 [37], and performing an enzyme assay designed to detect the phosphate ion released from GTP hydrolysis by Rab5b . Addition of the GAP domain markedly increased the GTP hydrolysis ( Figure 4B). In contrast, VipD had no effect on the GTP hydrolysis ( Figure 4B), demonstrating that VipD does not function as a GAP for Rab5b.

VipD abrogates the binding of downstream effectors to activated Rab5b and Rab22a
Another possibility was that VipD binding to activated Rab5b and Rab22a prevents the interactions with their direct down-stream effectors. Rabaptin-5 and Rabenosyn-5 bind directly to activated Rab5 and mediate endocytic membrane docking and fusion as well as early endosomal trafficking [38][39][40][41]. In a glutathione S-transferase (GST) pull-down assay, GST-tagged Rabaptin-5(739-862), encompassing the Rab5-binding domain of the protein [40], bound to the GTP-bound form, but not to the GDP-bound form of Rab5b ( Figure 5A; lanes 3 and 4). This complex was disrupted when VipD was challenged in a 1:1 molar ratio with GST-Rabaptin-5(739-862) ( Figure 5A; lane 5). Likewise, VipD disrupted the interaction between Rab5b(1-190;Q79L):GTP and GST-tagged Rabenosyn-5(1-70), which includes the Rab5-binding domain of the protein ( Figure 5B). We also examined whether VipD affects the interaction between Rab22a and its effector protein early endosome autoantigen 1 (EEA1), whose N-terminal C 2 H 2 Zn 2+ finger domain is necessary for binding Rab22a and for controlling endosomal trafficking [42,43]. VipD aptly displaced GST-EEA1(36-91) bound to   10 pull-down assay with full-length VipD. The indicated Rab proteins fused to CPD-(His) 10 were incubated with VipD and Co 2+ resin. Both the wild-type (WT) and the Q-to-L mutant Rabs were tested. Input proteins (I) and Co 2+ resin-bound proteins (R) were visualized on a denaturing gel. VipD bound to the active forms of Rab5a, Rab5b and Rab5c preferentially over the inactive forms of the three Rabs (A; lanes 2 to 5). VipD bound tightly to both the active and the inactive forms of the two isoforms of Rab22 (B; lanes 2 to 5). VipD did not interact with CPD-(His) 10 Figure 5C; lanes 3 to 5). Consistently with these in vitro displacement assays, the endogenous association between Rab5b and Rabaptin-5 in RAW264.7 macrophages was disrupted by the expression of full-length VipD or VipD(316-621), but not by the expression of VipD(1-316) ( Figure 5D). What would be the basis for the observed competitive binding of VipD to the activated Rabs? The Rab effectors commonly make contacts with a predominantly nonpolar surface of their cognate Rab, on which three highly conserved apolar residues (Phe57, Trp74 and Tyr89 in human Rab5b; see Figure S8) form a hydrophobic triad that is critical for the binding interaction [38,40,42,44,45]. In a (His) 10 pull-down assay, three Rab5b variants with an alanine substitution of one of the three residues exhibited no or barely detectable interaction with VipD ( Figure 5E), pointing that VipD also recognizes the hydrophobic triad and therefore competes with the effector molecules for binding to the activated Rabs. Together, these results indicate that the C-terminal domain of VipD is able to counteract the downstream signaling from the activated form of Rab5 and Rab22.

Blockade of endosomal trafficking by VipD
The capacity of VipD to disrupt the interactions between the three effectors and Rab5b or Rab22a strongly suggested that VipD interferes with endosomal trafficking leading to the degradation of endocytic materials. We therefore analyzed the effect of VipD expression on the transport and the degradation of exogenously added DQ-Red bovine serum albumin (BSA), which emits red fluorescence upon proteolytic degradation and is used as a sensitive indicator of lysosomal activity. In lipopolysaccharide (LPS)-treated RAW264.7 mouse macrophages, the degradation of DQ-Red BSA was significantly attenuated in cells stably expressing full-length VipD or VipD(316-621) compared with that in cells expressing vector alone or VipD(1-316) ( Figures 6A and 6B).   also blocked the degradation of phagocytosed E. coli in RAW264.7 cells, while the bacteria were disintegrated within 24 hours in macrophages expressing vector alone or VipD(1-316) ( Figure 6C). Next, time-course confocal microscopy was performed to identify which step of the endocytic degradation pathway was affected by VipD ( Figure 6D). LPS is recognized by the Toll-like receptor 4 (TLR4)-MD-2 complex and induces endocytic internalization and consequent lysosomal degradation of the receptor complex [46]. TLR4 was internalized into the RAW264.7 macrophage cytoplasm and colocalized with the early endosomal marker EEA1 within 20 min after LPS treatment, regardless of the expression of any VipD constructs ( Figure 6D; left panels), indicating that VipD does not interfere with the formation of endocytic vesicles or their heterotypic fusion with early endosomes. Critically, in 1 hour after LPS treatment, TLR4 colocalized with the late endosomal/ lysosomal marker lysosome-associated membrane protein-1 (LAMP-1) in cells expressing vector alone or VipD(1-316), but not in cells expressing full-length VipD or VipD(316-621) ( Figure 6D; right panels). These results suggest that VipD might block the endosome maturation step in macrophage cells via the C-terminal domain.
Stably expressed Rab5c was shown to be excluded from L. pneumophila-containing phagosomes in HeLa cells [47]. We sought to examine whether endogenous Rab5 might be excluded, and if it is, VipD might be responsible for the exclusion. C57BL/6 mouse bone marrow-derived macrophages (BMDM) and three different L. pneumophila mutant strains were prepared: Lp03 (dotA-deficient type IV secretion system-defective), DflaA (flagellin-gene deficient) and DvipD/DflaA (vipD and flaA-deficient). However, we found that endogenous Rab5b does not localize to the LCV in L. pneumophilainfected macrophages regardless of the strain background ( Figure  S10A; columns 1-3). In a positive control experiment, endogenous Rab1b localized to the LCV in cells infected by the DflaA or DvipD/DflaA strain but not in cells infected by the Lp03 strain ( Figure S10A; columns 4-6). Similar results were obtained with two different cell lines (macrophage-like human monocytic leukemia U937 and human alveolar basal epithelial A549 cells), which were infected by the L. pneumophila strain Lp02 (wild-type) or DvipD (vipD-deficient). In both type of cells, Rab5b did not localize to the LCV, irrespective of the presence of VipD (Figures S10B and S10C; rows 1-2). In contrast, Rab1b localized to the LCV in both types of cells infected by the Lp02 strain (Figures S10B and S10C; row 3). These observations reinforce the notion that Rab5 is excluded from the LCV, and suggest that at least VipD is not responsible for this exclusion. As expected, the Rab5 effectors EEA1 and Rabaptin-5 did not localize to the LCV, irrespective of the presence of VipD in these infected cells (Figures S10B and S10C; rows 4-7).

Discussion
L. pneumophila resides and replicates in macrophages, which is at the forefront against infectious agents. To understand L. pneumophila's strategies to evade the immune defense of macrophages, it is critical to know how pathogen's effector proteins manipulate host molecules. However, such information is yet very limited. Through elegant studies [6][7][8][9][10][11][12]17,[48][49][50][51], a number of L. pneumophila effectors, SidM/DrrA, SidD, LepB, AnkX and LidA, have been identified to target host Rab proteins, especially and commonly Rab1, a key regulator of ER-to-Golgi vesicle trafficking. Dysregulation of Rab1 by these effectors enables L. pneumophila to divert ER-derived vesicles to the LCV for the supply of nutrients and membrane components, highlighting that ER-to-Golgi vesicle trafficking is an important target for the intracellular growth of the pathogen. The study presented herein shows that endosomal vesicle trafficking is also targeted by L. pneumophila via VipD that blocks downstream signaling from Rab5 and Rab22. These two Rabs compose a Rab22-Rabex-5-Rab5 signaling relay [52], where activated Rab22 recruits Rabex-5, the GEF promoting the GDP-to-GTP exchange on Rab5 [28,29]. Activated Rab5 then recruits downstream effector proteins such as Rabaptin-5, Rabenosyn-5 and EEA1, which mediate diverse endosomal processes including vesicle fusion and membrane trafficking [39,41,53]. In addition, Rab22a regulates the formation of tubular recycling endosomes, which are necessary for endosome-to-plasma membrane recycling trafficking of internalized materials [31]. We show that VipD specifically and potently interacts with the two endosomal Rabs, blocking their binding interactions with the three downstream effectors through its C-terminal domain.
In the interaction of VipD with Rab5 and Rab22, three features are outstanding. First, VipD primarily targets the activated form of the two Rabs. Second, while activated Rab5 and Rab22 interact with their effector molecules weakly (K D .0.9 mM) [38,42], the binding affinity of VipD for these Rabs is exceedingly higher (K D ,254 nM). Third, VipD recognizes the conserved hydrophobic triad (Phe-Trp-Tyr), which is a common binding motif in diverse Rabs for the interaction with their downstream effector molecules [38,40,42,44,45]. These three features should enable VipD to potently block the downstream signaling from Rab5 and Rab22 by abrogating their association with the three effector molecules we tested in this study and probably with other effectors. To our knowledge, VipD is the first established example of a pathogen protein that antagonizes downstream signaling through binding to an activated Rab to competitively inhibit the binding of effector molecules. Of note, VipD does not interact with Rab7 ( Figure S7), which replaces Rab5 on early endosomes [54] and mediates endosomal-lysosomal trafficking [55]. VipD also does not interact with Rab4b, Rab9a, Rab14 and Rab21 ( Figure S7), which are known to mediate endosome-related trafficking [30]. Therefore, the observed endosomal trafficking block by VipD is most likely through selectively inhibiting the function of Rab5a, Rab5b, Rab5c and Rab22a.
In this study, we also confirmed that VipD has a phospholipase A 2 activity and that Ser73 and Asp288, invariant in cPLA 2 , VipD, VpdA, VpdB and ExoU [22], constitute a catalytic dyad in VipD ( Figures 1B and 1C). Since the N-terminal lipase domain of VipD is dispensable for VipD to localize to endosomes (Figures 2 and  S3), to bind Rab proteins ( Figure 3D) and to perturb endosomal trafficking (Figures 6 and S4B), the role of this domain is elusive. As VipD localizes to endosomes, one possibility is that VipD exhibits its catalytic activity on the endosomal membrane, the consequence of which remains to be elucidated.
In summary, the structural and biochemical analyses identified VipD as a signal blocker disabling the key endosomal regulators Rab5 and Rab22. As phagocytic vesicles could undergo fusion with lysosomes, our findings raise an important question of whether VipD facilitates the survival of L. pneumophila in macrophage, which needs further investigation. Our observations also form rational grounds for future investigations to delineate the role of the lipase activity of VipD and to decipher the functional roles of the Cterminal domain of the VipD-related bacterial effectors VpdA and VpdB, which are also translocated into host cells.

Crystallization and structure determination of VipD
The crystals of native VipD(1-575) were obtained by the hanging-drop vapor diffusion method at 22uC by mixing and The crystals of selenomethionine-substituted VipD(1-575) grew from a mixture of 100 mM MES (pH 6.0) and 1.3 M ammonium sulfate. Before data collection, the crystals were immersed in the precipitant supplemented with 30% glycerol and incubated overnight at 220uC. This dehydration process at high glycerol concentration improved the resolution of X-ray diffraction; from typical 5 Å up to 2.9 Å . The crystals were plunged into liquid nitrogen before X-ray data collection. X-ray data sets were collected using synchrotron X-ray radiation. The structure was determined by single-wavelength anomalous dispersion phasing using a selenomethionine-substituted VipD(1-575) crystal with the programs SHELX [56] and autoSHARP [57]. Subsequently, model building and refinement were carried out using the programs COOT [58] and CNS [59]. The final model does not include residues 559-575, whose electron densities were not observed or very weak. Crystallographic data statistics are summarized in Table 1.

Protein binding analysis
For (His) 10 pull-down assays, 25 mM of Rab-CPD-(His) 10 and 37.5 mM of VipD or VipD(316-621) were incubated at room temperature for 30 min and mixed with 30 mL of Co 2+ resin. The resin was washed four times with a buffer solution containing 20 mM Tris-HCl (pH 7.5), 100 mM NaCl and 2 mM MgCl 2 , and subjected to denaturing polyacrylamide gel electrophoresis. For quantification of protein-protein interaction, ITC measurements were carried out at 25uC on a microcalorimetry system iTC200 (GE Healthcare). Protein samples were prepared in a buffer solution containing 20 mM Tris-HCl (pH 7.5) and 100 mM NaCl. The samples were centrifuged to remove any residuals prior to the measurements. Dilution enthalpies were determined in separate experiments (titrant into buffer) and subtracted from the enthalpies of the binding between the proteins. Data were analyzed using the Origin software (OriginLab).

Cell-based assay
For the subcellular localization analysis, HeLa cells and mouse macrophage RAW264.7 cells were transfected with the pEYFP-N1 or pECFP-C1 vectors (Clontech) encoding Rab or VipD proteins and visualized by confocal microscopy. For the analysis of endocytic trafficking, RAW264.7 cells were transfected with the pCDH-CMV vector (System Biosciences) encoding VipD proteins, and stable cell lines were established by puromycin selection. The details of mammalian cell culture, immunoblotting, flow cytometry and live cell imaging are described in Text S1.

Accession codes
The coordinates of the VipD(1-575) structure together with the structure factors have been deposited in the Protein Data Bank with the accession code 4AKF.  Figure S6 RNA interference assay. HeLa cells transiently expressing YFP-tagged VipD or CFP-tagged Rab proteins were indicated VipD proteins were fixed at 20 min (left) or 1 h (right) after LPS treatment. TLR4, EEA1 and LAMP-1 were immunostained and visualized. TLR4 internalization was clearly observable in LPS-treated cells but not in untreated cells (see Figure S9). The internalized TLR4 colocalized with EEA1, but not with LAMP-1 in cells expressing full-length VipD or VipD(316-621). The LAMP-1 signal was consistently weak in cells expressing full-length VipD or VipD(316-621) (column 5; rows 2 and 3). The reason is unclear, but it may be due to the blockade of endosomal maturation by the VipD proteins, which could cause attenuated LAMP-1 localization to late endosomes. The scale bars indicate 5 mm. doi:10.1371/journal.ppat.1003082.g006 treated with the indicated siRNAs and visualized by confocal microscopy. The treatment of siRNA blocked the expression of the target Rab proteins (first and second rows). The endosomal localization of VipD was not affected by the siRNA treatment (third row). The scale bar indicates 10 mm. (TIF) Figure S7 (His) 10 pull-down assay. Full-length VipD and each of the indicated GDP-bound (top) or GTP-bound (middle) Rabs fused to CPD-(His) 10 were incubated together with Co 2+ resin, and a (His) 10 pull-down assay was performed as in Figure 3. None of the Rabs exhibited a notable coprecipitation with VipD except Rab22a used as a control. The table lists the Rab proteins tested in the pull-down assay and the intracellular trafficking they are involved in. (TIF) Figure S8 Sequence alignment of Rabs. The twelve different Rab proteins presented in this manuscript are aligned. The highly conserved three nonpolar residues commonly involved in binding to host effectors and to VipD are indicated by asterisks. Text S1 Details of protein preparation, biochemical assays and cellular assays. (DOC)