Report

A Bifunctional Bacterial Protein Links GDI Displacement to Rab1 Activation

See allHide authors and affiliations

Science  09 Nov 2007:
Vol. 318, Issue 5852, pp. 974-977
DOI: 10.1126/science.1149121

Abstract

Rab guanosine triphosphatases (GTPases) regulate vesicle trafficking in eukaryotic cells by reversibly associating with lipid membranes. Inactive Rab GTPases are maintained in the cytosol by binding to GDP-dissociation inhibitor (GDI). It is believed that specialized proteins are required to displace GDI from Rab GTPases before Rab activation by guanosine diphosphate–guanosine 5′-triphosphate (GDP-GTP) exchange factors (GEFs). Here, we found that SidM from Legionella pneumophila could act as both GEF and GDI-displacement factor (GDF) for Rab1. Rab1 released from GDI was inserted into liposomal membranes and was used as a substrate for SidM-mediated nucleotide exchange. During host cell infection, recruitment of Rab1 to Legionella-containing vacuoles depended on the GDF activity of SidM. Thus, GDF and GEF activity can be promoted by a single protein, and GDF activity can coordinate Rab1 recruitment from the GDI-bound pool.

Vesicle trafficking between membrane-bound organelles in eukaryotic cells is regulated by the highly conserved Rab family of small guanosine triphosphatases (GTPases) (1). Rab proteins activated by guanosine diphosphate–guanosine 5′-triphosphate (GDP-GTP) exchange factors (GEFs) are associated with membranes by means of two hydrophobic prenyl groups covalently attached to their C terminus (26). Inactivated GDP-Rabs are extracted from membranes and maintained in the cytosol by GDP-dissociation inhibitor (GDI) (711). GDI binding prevents spontaneous GDP-GTP exchange on Rab proteins, and, therefore, their inappropriate activation (12).

Delivery of Rab GTPases to sites of activation involves their release from GDI, a process thought to require GDI-displacement factors (GDFs) (13, 14). PRA-1 (the human homolog of Yip3) facilitates dissociation of endosomal Rab proteins from GDI and represents a protein with GDF activity discrete from a GEF (15). The identity and function of other proteins with GDF activity has remained elusive.

The human pathogen Legionella pneumophila establishes a replication vacuole within alveolar macrophages by recruiting material from the host cell endoplasmic reticulum (ER) (1618). The bacterium translocates a large cohort of effector proteins into the host cell cytosol by means of its Dot-Icm protein type IV secretion system (T4SS) (19). Two of these translocated proteins, SidM (DrrA) and LidA, target mammalian Rab1 (20, 21), the key regulator of endoplasmic reticulum (ER) to Golgi vesicle transport (22, 23). SidM was necessary for the recruitment of Rab1 to the Legionella-containing vacuole (LCV) and had GEF activity specifically toward Rab1 (20, 21). In vitro, surface-immobilized SidM and LidA could collaborate to tether ER-derived vesicles dependent on activated Rab1 (20), which suggests that both L. pneumophila effectors are crucial for intercepting Rab1-controlled secretory vesicle transport during infection.

Although recruitment of host cell Rab1 to the LCV required SidM (20, 21), the Rab1-GEF activity of SidM was unlikely to be responsible for this capturing event, because Rab1 must be released from GDI before nucleotide exchange can occur. The L. pneumophila genome does not encode any obvious PRA1 homologs that could trigger GDI dissociation from Rab1. Thus, we tested whether SidM and/or LidA had GDF activity toward the Rab1-GDI complex. Because GDI preferentially binds prenylated Rab proteins (24), and bacteria lack the enzymatic machinery for Rab prenylation, Saccharomyces cerevisiae was used for the production of prenylated hexahistidine (His6)-tagged human Rab1B (ScRab1) and FLAG-tagged human GDI2 (ScGDI2) (fig. S1). Double prenylation of purified ScRab1 was confirmed (25), and purified ScRab1 and ScGDI2 formed a heterodimeric 1:1 complex (fig. S2). When ScRab1-ScGDI2 was incubated with bead-immobilized recombinant SidM or LidA, ScRab1 only bound to SidM-coated beads but not to LidA-coated beads or uncoated control beads (Fig. 1A). ScGDI2 did not coprecipitate with SidM-bound ScRab1, which indicated that it was displaced from the complex. Nonprenylated GDI-free human Rab1 purified from Escherichia coli (EcRab1) was bound equally by SidM- and LidA-coated beads (Fig. 1A) (20). Thus, the Rab1-GEF SidM exhibited GDF activity and promoted dissociation of the ScRab1-ScGDI2 complex during ScRab1 binding, whereas LidA binding was restricted to the GDI-free form of Rab1.

Fig. 1.

Rab1 activation by SidM involves GDI displacement and membrane insertion of ScRab1. (A) ScRab1 pull-down assay shows activity by SidM but not LidA. Bead-immobilized SidM or LidA or uncoated control beads were incubated with ScRab1-ScGDI2 (top panel) or EcRab1 (bottom panel), and proteins on the beads (bound) or in the supernatant (unbound) were detected by immunoblot analysis using antibody against GDI2 or Rab1B. (B and C) For a [γ-35S]GTP incorporation experiment, we used 2 pmol SidM and either 40 pmol ScRab1-ScGDI2 (B) or 20 pmol EcRab1 (C). [γ-35S]GTP incorporation was detected as the increase in radioactivity over time. (D) [γ-35S]GTP incorporation by SidM (1.6 pmol) into a molar excess of ScRab1-ScGDI2 (10 pmol) is accelerated in the presence of phosphatidylcholine (PC) liposomes. The curves of nucleotide-exchange studies in the absence of PC vesicles are shown with another scale in the inset. (E) Protein association with PC vesicles during nucleotide exchange. ScRab1-ScGDI2 was incubated with PC liposomes in the presence or absence of SidM or GTP-γ-S (as indicated). Liposomes were separated from soluble proteins by sucrose-gradient centrifugation, and proteins in the vesicle fraction were detected by immunoblot analysis. (F) Analysis of LidA association with PC vesicles previously incubated with ScRab1-ScGDI2 in the presence or absence of SidM and GTP-γ-S (as indicated). (A) to (F) represent at least two repetitions.

Because SidM dissociated Rab1 from GDI (Fig. 1A), we tested whether SidM could catalyze exchange of GDP against radiolabeled [γ-35S]GTP-γ-S [guanosine 5′-O-(3′-thiotriphosphate, a nonhydrolyzable GTP analog] in ScRab1 in complex with ScGDI2. In the absence of SidM, ScRab1-ScGDI2 showed only minimal spontaneous [γ-35S]GTP uptake (Fig. 1B). In contrast, addition of SidM led to the rapid incorporation of [γ-35S]GTP into ScRab1. However, the amount of [γ-35S]GTP incorporated into ScRab1 in complex with ScGDI2 (Fig. 1B) was consistently lower than that observed for EcRab1 (Fig. 1C). Because active Rab GTPases associate with the membrane by means of their C-terminal prenyl groups in vivo (5, 6, 14, 2628), the lack of a lipid bilayer during the in vitro GDP–[γ-35S]GTP–exchange studies (Fig. 1B) may explain the attenuated activation of prenylated ScRab1. Consistent with this, we found that SidM-promoted [γ-35S]GTP incorporation into ScRab1 in the presence of phosphatidylcholine (PC) liposomes was more than four times that seen in the absence of lipid vesicles (Fig. 1D). Thus, ScRab1 activation by SidM was kinetically enhanced in the presence of a lipid bilayer, presumably because the bilayer provided a hydrophobic environment into which GDI-free ScRab1 could be incorporated by SidM.

Next, we performed vesicle floating experiments to monitor protein association with liposomes during the different stages of nucleotide exchange. Incubation of SidM with a molar excess of ScRab1-ScGDI2 revealed that accumulation of ScRab1 on PC vesicles occurred only in the presence of both SidM and GTP-γ-S (Fig. 1E). Without GTP-γ-S, only a fraction of ScRab1 was integrated into PC vesicles and remained bound by SidM, which showed that nucleotide exchange was required to allow release of membrane-bound ScRab1 by SidM before another ScRab1-ScGDI2 complex was targeted. When LidA was added to PC liposomes, the protein specifically interacted with vesicles enriched in GTP-γ-S-ScRab1 (Fig. 1F). Membrane-bound ScRab1 activated and released by SidM was thus available to bind downstream acceptors such as LidA. The SidM-mediated Rab1 activation process can therefore be divided into three stages: (i) GDI release from Rab1 accompanied by membrane insertion of SidM-bound Rab1, (ii) catalysis of nucleotide exchange in Rab1, and (iii) release of activated Rab1 by SidM to allow downstream ligand binding to GTP-Rab1. Alternatively, SidM could have indirectly accelerated the dissociation of ScGDI by catalyzing nucleotide exchange in ScRab1 while bound to ScGDI, thereby generating GTP-charged ScRab1 that is released by ScGDI.

The discovery of both GEF and GDF activity within the same polypeptide chain was surprising. Thus, we wanted to map the regions of SidM contributing to each of these activities. Purified SidM variants with N- or C-terminal truncations (Fig. 2A and fig. S3) were analyzed for their ability to interact with EcRab1. Deletions exceeding amino acids 1 to 316 at the N terminus or residues 545 to 647 at the C terminus disrupted binding of SidM variants to EcRab1 (Fig. 2B). In contrast, fragments containing amino acids 317 to 545 bound EcRab1 as efficiently as full-length SidM. Similar results were obtained in pull-down studies using EcRab1(S25N) (Fig. 2A), a GDP-locked mutant (29). Thus, the central region of SidM containing amino acids 317 to 545 was required for EcRab1 binding.

Fig. 2.

Domain mapping of SidM. (A) (Left) Schematic representation of SidM variants used in this study (numbers indicate amino acid residues). (Right) Summary of the experiments shown below in (B to E) indicating positive (+) or negative (–) outcomes. The central region between amino acid residues 317 and 545 (shaded blue) was found to be essential for all in vitro activities of SidM. (B and C) Pull-down of EcRab1 (B) or ScRab1-ScGDI2 (C) by bead-immobilized SidM variants. Proteins bound to the beads or unbound in the supernatant were detected by immunoblot analysis (using FLAG-specific or Rab1B-specific antibody). The lane marked with an asterisk shows a 28.1-kD protein cross-reacting with an antibody [corresponding to the bait protein SidM(396–647)]. (D and E) [γ-35S]GTP incorporation into (D) EcRab1 [(left) 20 pmol, (right) 10 pmol] or (E) ScRab1-ScGDI2 (10 pmol) by SidM variants (2 pmol). Rab1 proteins and SidM variants were incubated, and [γ-35S]GTP uptake by Rab1 proteins was determined as the increase in radioactivity over time. (B) to (E) represent at least two repetitions.

Similarly, we found that only SidM variants containing amino acid residues 317 to 545 associated with ScRab1 (Fig. 2C). ScGDI2 did not precipitate with ScRab1 on the beads and remained in the supernatant (Fig. 2C), consistent with its exclusion from the complex of ScRab1 bound to SidM variants.

To identify the GEF domain, SidM variants with N- or C-terminal truncations were analyzed for their ability to mediate incorporation of [γ-35S]GTP into EcRab1. Only SidM fragments containing the entire region required for Rab1 binding (residues 317 to 545) showed GEF activity equivalent to full-length SidM (Fig. 2D). Likewise, only those SidM variants that had GEF activity toward EcRab1 catalyzed GDP–[γ-35S]GTP exchange when incubated with ScRab1-ScGDI2, with efficiencies comparable to full-length SidM (Fig. 2E). Thus, GDI displacement, Rab1 binding, and nucleotide exchange activity required the same central region within SidM (Fig. 2A).

We also characterized a SidM variant consisting of only amino acid residues 317 to 545. Purified recombinant SidM(317–545) efficiently bound to immobilized nonprenylated glutathione S-transferase (GST)–tagged EcRab1 (Fig. 3A). Similarly, SidM(317–545) was capable of dissociating ScGDI2 from ScRab1 and displayed ScRab1-binding activity comparable to that of full-length SidM (Fig. 3B). Furthermore, SidM(317–545) efficiently catalyzed GDP–[γ-35S]GTP exchange in both EcRab1 (Fig. 3C) and ScRab1-ScGDI2 (Fig. 3D). Thus, the central region of SidM comprising amino acid residues 317 to 545 was sufficient to mediate the activities of GDI displacement, Rab1 binding, and nucleotide exchange.

Fig. 3.

The central region of SidM is sufficient for GDI displacement, Rab1 binding, and GEF activity. (A) Immunoblot showing precipitation of SidM(317–545) by GST-Rab1–coated beads but not by GST-bound control beads. Lane marked with an asterisk shows a protein of 25 kD (corresponding to the bait protein GST) cross-reacting with the antibody. (B) ScGDI2 displacement and ScRab1 binding by SidM (317–545). ScRab1-ScGDI2 was incubated with bead-immobilized SidM variants (numbers indicate amino acid residues) or uncoated control beads, and proteins on the beads (bound) or in the supernatant (unbound) were detected by immunoblot analysis. (C and D) SidM(317–545) has GEF activity. EcRab1 (10 pmol) (C) or ScRab1-ScGDI2 (12.5 pmol) (D) were incubated with SidM variants (3 pmol), and uptake of [γ-35S]GTP by Rab1 protein was determined over time. All figure parts represent at least two repetitions.

Both SidM and LidA efficiently interacted with Rab1 and localized to the cytoplasmic surface of the LCV (20, 30). Nevertheless, only SidM was required for the recruitment of endogenous Rab1 to LCVs (fig. S4) (20, 21). Because SidM dissociated the ScRab1-ScGDI2 complex in vitro (Fig. 1A), we asked whether GDF activity allowed SidM to target the cytoplasmic pool of GDI-bound host cell Rab1, and whether the lack of GDF activity in LidA prevented this effector from recruiting Rab1 complexed by GDI. Thus, we tested whether the requirement for SidM during Rab1 recruitment to LCVs could be bypassed by LidA in the presence of Rab1(D44N) (fig. S5), a Rab1 mutant in which Asn (N) replaces Arg (D) at residue 44 and that shares all features described for wild-type Rab1 but that is unable to bind GDI (31). The D44N substitution in Rab1 did not interfere with LidA binding, as GST-Rab1(D44N) and GST-Rab1 purified from E. coli were bound equally by recombinant LidA (Fig. 4A). When COS-1 cells producing green fluorescent protein (GFP)-tagged Rab1(D44N) were infected with L. pneumophila (Fig. 4, B and C), GFP-Rab1(D44N) was efficiently recruited to LCVs containing wild-type (44 ± 8%) but not the type IV secretion–defective (T4SS) strain Lp03 (12 ± 3%), which indicated that GFP-Rab1(D44N) recruitment was dependent on a functional Dot-Icm transporter. L. pneumophila DsidM, which was defective for the recruitment of wild-type Rab1 (20, 21), showed robust colocalization with GFP-Rab1(D44N) (51 ± 10%). This SidM-independent recruitment of GFP-Rab1(D44N) was mediated by LidA, as deletion of both sidM and lidA in L. pneumophila abrogated recruitment of GFP-Rab1(D44N) to LCVs (17 ± 7%). Thus, LidA binding in vivo was restricted to GDI-free Rab1, which explains why release of GDI and recruitment of Rab1 from the cytoplasmic pool required the GDF activity of SidM. The inability of LidA to extract Rab1 from GDI indicated a hierarchy between these two L. pneumophila effectors, presumably to ensure that Rab1 recruited to the LCV was activated by SidM before its association with LidA, which assists in accumulating activated Rab1 about the LCV. Indeed, vacuoles containing L. pneumophila ΔlidA showed delayed Rab1 recruitment compared with wild-type LCVs (20).

Fig. 4.

LidA can recruit GDI-free Rab1(D44N) in the absence of SidM. (A) LidA can bind Rab1(D44N). Binding of LidA to bead-immobilized GST-Rab1 or GST-Rab1(D44N) but not GST was determined by SDS–polyacrylamide electrophoresis and Coomassie staining. The figure represents two repetitions. (B) Recruitment of GFP-Rab1(D44N) during L. pneumophila infection. Transiently transfected COS-1 cells producing GFP-Rab1(D44N) were infected for 30 min as indicated. Cells were fixed and stained for intracellular bacteria (left). (Middle) GFP-Rab1(D44N), (right) merged images with bacteria (red) and GFP-Rab1(D44N) (green). Arrows indicate the location of the LCV magnified threefold in the insets of each panel. Scale bar, 5 μm. Contrast was equally changed by linear adjustment. (C) Quantification of (B) showing that recruitment of GFP-Rab1(D44N) in the absence of SidM requires LidA. The graph represents pooled data (mean ± SD) from four independent experiments. *P = 0.0001 (Student′s t test).

SidM is a protein that has both GEF and GDF activity toward a Rab GTPase. This unique ability of SidM to link GDI displacement to Rab1 activation explains how the intravacuolar pathogen L. pneumophila can efficiently exploit host cell Rab1 even in the presence of GDI that naturally interferes with this process. The discovery of both GEF and GDF activity within SidM raises the intriguing possibility that eukaryotic GEF proteins may possess similar abilities to mediate membrane delivery and activation of Rab GTPases during intracellular vesicle transport.

Supporting Online Material

www.sciencemag.org/cgi/content/full/1149121/DC1

Materials and Methods

Figs. S1 to S5

Table S1

References

References and Notes

View Abstract

Navigate This Article