Salmonella infections require the delivery of bacterial effectors into the host cell that alter the regulation of host defense mechanisms. The secreted cysteine protease GtgE from S. Typhimurium manipulates vesicular trafficking by modifying the Rab32 subfamily via cleaving the regulatory switch I region. Here we present a comprehensive biochemical, structural, and computational characterization of GtgE in complex with Rab32. Interestingly, GtgE solely processes the inactive GDP-bound GTPase. The crystal structure of the Rab32:GDP substrate in complex with the inactive mutant GtgEC45A reveals the molecular basis of substrate recognition. In combination with atomistic molecular dynamics simulations, the structural determinants for protein and activity-state specificity are identified. Mutations in a central interaction hub lead to loss of the strict GDP specificity. Our findings shed light on the sequence of host cell manipulation events during Salmonella infection and provide an explanation for the dependence on the co-secreted GTPase activating protein SopD2.
Central regulatory elements of intracellular signaling networks are small GTPases (guanosine triphosphate hydrolases, also referred to as G-proteins). G-proteins function as binary molecular switches and thereby inhibit or stimulate cellular networks via spatio-temporal protein–protein interactions1. In order to function in this capacity, the small GTPases can exist in the following two states: (1) they are inactive when complexed to guanosine diphosphate (GDP), and (2) they become active once bound to guanosine triphosphate (GTP). In the active state, signals are propagated by the recruitment of effector proteins that function as signal transmitting molecules. The activity states are regulated by dedicated regulatory enzymes that facilitate activation and inactivation of theG-proteins: guanine nucleotide exchange factors (GEFs) activate small GTPases by stimulating GDP-release and facilitating GTP-binding. In contrast, GTPase activating proteins (GAPs) accelerate inactivation via increasing the rate of intrinsic GTP-hydrolysis, converting the G-protein back to the inactive state. The activity states of small GTPases are manifested in specific conformations of two highly important regulatory loop regions that are referred to as switch I and switch II: the switch-regions are structurally disordered in the inactive state, but adopt a well-defined structure in the active form. The Rab-family of small GTPases is involved in regulating intracellular vesicular trafficking temporally and spatially1. These proteins are part of the so-called Rab-cycle that is characterized by a series of peripheral membrane association and cytosolic localization linked to nucleotide exchange. In order to attach to a membrane, Rab-proteins are post-translationally modified with one or two geranylgeranyl lipids at their C-termini. In the active state, these lipids confer strong membrane affinity and localize the Rab-protein to the cytosolic surface of intracellular compartments. In the inactive state, however, lipidated Rabs are targeted with high affinity by the GDP-dissociation inhibitor (GDI) that solubilizes the GTPase and thereby causes cytosolic localization.
Due to the pivotal role of small GTPases in intracellular signaling and the maintenance of cellular homeostasis, they are frequently targeted by bacterial pathogens in order to promote the infection2. An intriguing example in this respect is the bacterium Salmonella enterica. This bacterial pathogen is composed of a number of serovars that can infect diverse vertebrate species3,4. The Salmonella enterica Typhi (S. Typhi) causes typhoid fever by infecting only humans, whereas the Salmonella enterica Typhimurium (S. Typhimurium) can infect a broad host range. The molecular basis of this host range selection is not entirely clear. Recently, however, Galán and coworkers discovered that a protein encoded by the gene gtgE in S. Typhimurium contributes to host selection5. Consequently, expression of gtgE in S. Typhi counteracts host restriction and enables colonization of mice, thus, broadening host specificity. The protein GtgE has been identified as a cysteine protease that targets members of the Rab-family5,6,7,8. In particular, action of GtgE results in specific proteolytic cleavage of Rab29, Rab32, and Rab38 in vitro. The cleavage site is located in the regulatory switch I region of these Rabs and it takes place between G59 and V60 (amino acid numbering refer to Rab32)6. Rab32 is involved in controlling the biogenesis of lysosome-related organelles (LRO)9,10,11,12,13,14. It is believed that GtgE inactivates Rab32 proteolytically and thereby prevents the delivery of antimicrobial factors that would impair Salmonella infection. Since Salmonella is taken up by phagocytosis and evades destruction in the Salmonella-containing vacuole (SCV), which also constitutes an LRO, the Rab32-inactivating activity of GtgE may be an important component of the infectious mechanism5.
Individual atomic structures of GtgE have been reported recently7,8. However, the molecular basis of the recognition of Rab32 by GtgE and the biochemical details of binding have not yet been addressed. Here we present the crystal structure of the Rab32:GtgE complex. Furthermore, by using a combination of biochemical characterization, molecular dynamics simulation, and binding studies, we demonstrate that GtgE shows exclusive selectivity for the inactive GDP-bound state of Rab32. The biochemical characterization of the GDP-state specificity of GtgE explains the previously reported necessity of Rab32:GTP inactivation by the secreted Salmonella GAP SopD215.
GtgE specifically cleaves inactive, GDP-bound GTPases
GtgE is a cysteine protease that cleaves the Rab32-subfamily comprising Rab29, Rab32, and Rab38, but not the closest related homolog Rab235. Initially, we verified the previously observed Rab-specificity of GtgE (Fig. 1a). GtgE-mediated Rab-cleavage was monitored using denaturing sodium dodecylsulfate polyacrylamide gel electrophoresis (SDS-PAGE) on the basis of the molecular weight decrease by 6.5 kDa (Rab32, proteolysis at G59) or by 4.5 kDa (Rab29, Rab38, proteolysis at G41 and G43, respectively) (for sequence alignment of Rab32, Rab29, Rab38 and Rab23, Supplementary Fig. 1). Indeed, Rab29, Rab32, and Rab38, but not Rab23, are cleaved in the presence of catalytic quantities of GtgE.
We next asked whether GtgE is selective for Rab-proteins inthe active (GTP) or inactive (GDP) state. To this end, the Rab-proteins were prepared in defined nucleotide states with GDP, GTP, or the non-hydrolyzable GTP-analog GppNHp. SDS-PAGE based monitoring of GtgE-mediated Rab-cleavage indicates that only the inactive GDP-bound, but not the active GppNHp-loaded GTPases, are processed by GtgE (Fig. 1b). Additionally, active Rab32 was purified from expression in Escherichia coli bound to GTP to ~90% (Supplementary Fig. 2A). Upon incubation with GtgE, Rab32:GTP was proteolytically cleaved significantly slower than the GDP-loaded form (Fig. 1b aand Supplementary Fig. 2B). In order to investigate whether Rab32:GTP or the Rab32:GDP produced via intrinsic GTP-hydrolysis are the target of GtgE, we separately quantified GtgE-mediated Rab32:GTP-proteolysis and GTP-hydrolysis. Rab32 has an exceptionally slow intrinsic GTPase-activity with an hydrolysis constant determined to k hydr = 6.5·10−6 s−1 (Fig. 1c, Supplementary Fig. 2A). The rate of Rab32:GTP-proteolysis in the presence of GtgE is identical to the rate of GTP-hydrolysis with an observed rate constant of k obs,prot = 6.7·10−6 s−1 (Fig. 1c, Supplementary Fig. 2B). This demonstrates that the proteolytic turnover of Rab32:GTP is limited by the rate of intrinsic GTP-hydrolysis. Consequently, only the inactive (GDP-bound), but not the active (GTP-bound) Rab, is the target of GtgE.
We therefore wondered whether GtgE cooperates with other enzymes in order to convert the Rab-proteins into the inactive form as required for proteolysis. Interestingly, Salmonella secretes the bacterial Rab32-GAP SopD2 that produces Rab32:GDP from Rab32:GTP15. Indeed, proteolysis only happens in vitro when SopD2 is added catalytically to the reaction between Rab32:GTP and GtgE (Supplementary Fig. 3). The rate of GtgE-mediated cleavage for Rab32:GTP in the presence of SopD2 was indistinguishable from the reaction of Rab32:GDP in the absence of this GAP. This indicates a cooperative deactivation mechanism of SopD2 with GtgE for effective Rab32-cleavage.
We further note that Rab32:GDP-proteolysis in vitro does not precipitate the protein although the cleavage site is located in the structurally and functionally crucial switch I region (Supplementary Fig. 4). In fact, GtgE-cleaved Rab32:GDP could be preparatively produced and purified by size-exclusion chromatography. The protein elutes indistinguishably from the uncleaved form (Fig. 1d). We thus probed whether switch I cleavage could affect the folding stability of the G-protein. Rab32:GDP and GtgE-cleaved Rab32:GDP were subjected to thermal unfolding experiments using differential scanning fluorimetry (DSF) (Fig. 1e). The melting points (T m) were determined from fitting the fluorescence data to Boltzmann equation to T m,FL = 52 °C and T m,cleaved = 48 °C for uncleaved and cleaved Rab32:GDP, thus revealing that cleaved Rab32:GDP behaves like a well-folded and stable protein. In summary, we have shown that GtgE demonstrates exclusive specificity for inactive GDP-bound Rab-proteins, and that the cleaved Rab32 remains a stable monomeric protein.
GtgE has high affinity for inactive Rab substrates
We next addressed the enzymatic properties of GtgE for inactive Rab32:GDP. First, time-resolved kinetics of GtgE-mediated Rab32:GDP cleavage were performed. We processed Rab32:GDP with catalytic quantities of GtgE, separated heat-inactivated samples from different reaction time points on the SDS-PAGE, and quantified Rab32-cleavage educts by densitometric quantification of Coomassie-stained bands since traditional fluorescence-based approaches did not function in this case (Fig. 2a, b and see Supplementary Fig. 5A, B). The catalytic efficiency (k cat/K m) of GtgE was determined to 5·105 M−1 s−1. For this purpose, the observed rate constant (k obs) was obtained from fitting the dependence of Rab32FL-cleavage over time to a single exponential curve (Fig. 2b), followed by division of k obs by the GtgE concentration. The obtained k cat/K m value for GtgE implies a high catalytic activity, and suggests that GtgE operates at the maximum proteolysis rate even at low micromolar concentrations of Rab32:GDP. This property is indicative of a high affinity enzyme-substrate-complex having a Michaelis–Menten constant (K M) in the micro- or nanomolar range. Determination of the K M would require Michaelis–Menten-kinetics to be conducted at micromolar to submicromolar Rab32:GDP-concentrations. However, at such low substrate concentration, the quantification of cleavage educts and products by Coomassie-stained SDS-PAGE gels would be challenging. As an alternative, we aimed to determine the dissociation equilibrium constant (K D) between Rab32:GDP and a catalytically inactive GtgE-mutant as a surrogate for K M. The amino acid residue C45 has previously been shown to be important for the proteolytic activity of GtgE, and likely constitutes the active site nucleophile of this cysteine protease7,8. We therefore confirmed that GtgEC45A indeed has no proteolytic activity on Rab32:GDP (Supplementary Fig. 6), allowing us to determine the K D by two independent approaches. First, we subjected Rab32:GDP (20 µM) to isothermal titration calorimetry (ITC) with GtgEC45A (200 µM), revealing a dissociation constant of K D,ITC = 96 ± 31 nM (Fig. 2c). The determined stoichiometry confirms that GtgE and Rab32:GDP form a heterodimeric 1:1-complex. In parallel, we utilized analytical ultracentrifugation (aUC) of Rab32:GDP (200 nM) covalently labeled at thiol groups with Atto488-Maleimide (Rab32:GDPfluor.), and quantified complex formation with increasing concentrations of GtgEC45A by the difference in Rab32:GDPfluor. sedimentation (Fig. 2d). Assuming a 1:1 binding stoichiometry, the K D of the interaction is determined to K D,aUC = 118 ± 29 nM which is in good agreement with the ITC-experiment. Thus, our orthogonal approaches determined the K D for the interaction between Rab32:GDP and the inactive GtgEC45A mutant to be approximately 100 nM. Assuming that the K D equals the Michaelis–Menten constant K m for the enzymatic process, we obtain a turnover rate(k cat) of k cat = 0.05 s−1 by multiplying k cat/K m (Rab32:GDP, GtgE) = 5.1·105 M−1 s−1 with the K D. In summary, as indicated bya low K D and a moderate k cat, our findings indicate that GtgE is a catalytically efficient enzyme with a high substrate affinity.
The crystal structure of the Rab32:GDP:GtgE complex
The high affinity of GtgEC45A for Rab32:GDP allowed us to purify thecomplex (Supplementary Fig. 7). Crystals obtained after reductively alkylating the truncated Rab3218-201:GDP:GtgE21-214,C45A-complex and the full-length complex Rab321-225:GDP:GtgE1-228,C45A diffracted up to 2.3 Å and 2.9 Å resolution, respectively (Supplementary Fig. 8)16. The X-ray structures of the complexes were solved by the molecular replacement technique using the coordinates of Rab32:GppCH2p (PDB ID: 4CYM17) and truncated GtgE (PDB ID: 4MI77) as appropriate search models (Fig. 3a) (see Supplementary Table 1 for data collection and refinement statistics). The average B-factor of the full-length complex is higher than for the truncated complex, though the entire electron density is well defined for both structures (Supplementary Fig. 9). Due to the higher resolution, the truncated Rab3218-201:GDP:GtgE21-214,C45A-complex is discussed only and referred to as Rab32:GDP:GtgEC45A hereafter.
The complex shows the general features of the previously determined individual subunits of Rab32 and GtgE7,8,17. The Cα-atoms of the Rab32-molecules superimpose for 121 amino acids (excluding the structurally-flexible switch-regions, switch Iaa48-61 and switch IIaa82-99) with an RMSD of 0.52 Å18. In case of GtgE, 97 Cαs were aligned with an RMSD of 0.65 Å (excluding the previously unresolved amino acids 23–54). Rab32 is composed of a central β-sheet comprising six β-strands (β1R-β6R, where subscript R denotes Rab32) and five α-helices (α1R-α5R) (Fig. 3a, b; Supplementary Fig. 1). However, no electron density is observed for the helix α5R, which is usually present in all small GTPases. GtgE is composed of a six-membered central antiparallel β-sheet that is perpendicularly arranged in a tip-to-tip manner to the β-sheet of Rab32. The GtgE-β-sheet is flanked by five α-helices of which the two N-terminal α-helices and the three C-terminal helices are on opposite ends of the sheet structure (Fig. 3c)8. The amino acid regions 23–44 and 60–82 appear to be folded within the active enzyme structure only after complex formation with Rab32, as these peptide stretches occupy significantly different positions in comparison to the GtgE-apo-structure (Fig. 3c).
In the complex structure, GtgE can be envisioned as a trident poking into Rab32 with three spike-like protrusions (Fig. 3a, d). Spike 1 comprises amino acid region 38–45 in conjunction with region 76–82. Spike 2 is represented by the β2G–β3G loop (where subscript G denotes GtgE), and penetrates deeply into the switch I–switch II cleft of Rab32. Thereby, spike 1 and 2 form intimate interactions with the switch I structure that is also the site of proteolytic cleavage. Spike 3 is formed by the α4G-helix and a C-terminal loop extension. The shallow cleft between spike 2 and 3 is filled by the switch II region of Rab32.
Binding of GtgE to Rab32 does not result in major structural changes outside of the switch-regions when compared to the structure of Rab32:GppCH2p (Fig. 3b)17. Since GtgE is specific for the inactive GDP-bound Rab32 (Fig. 1), this suggests that the contacts within switch-regions undergo large-scale conformational changes upon GtgE-binding. The switches are reminiscent of, but not identical to the conformations observed in Rab-proteins in complex with the GDP-state-specific interaction partner GDI19,20 (Supplementary Fig. 10). Spike 2 of GtgE acts like a wedge that drives the switch-regions further apart than observed in Rab:GDI-complexes. On the other hand, binding of Rab32 to GtgE does not affect the secondary structure elements of the protease (Fig. 3c)7,8. Rather, flexible loops previously unrecognized inthe GtgE-structure, become ordered in the Rab32:GtgE complex. In particular, the N-terminus (amino acids 23–45), including the position of the active sitecysteine (here the C45A substitution) is now resolved, in addition to the loop regions β2G-β3G (i.e., spike 2) and α4G-β6G (i.e., spike 3).
The complex interface buries an average solvent exposed surface area of 1535 Å2, as determined with the ‘Protein interfaces, surfaces and assemblies’ web service (PISA)21. It is mainly composed of the switch- and β2R-β3R-regions of Rab32 and the three spikes of the GtgE-trident (Fig. 3d, e). A large number of hydrophobic and polar interactions contribute to the complex interface, providing a potential molecular basis for the observed high affinity between GtgE and Rab32:GDP (Fig. 3e). In particular, the Rab32 residues adjacent to the cleavage site between G59R and V60R are intimately involved in interacting with the deep spike 1/spike 2 cavity of GtgE. The switch I residue I58R forms the center of a hydrophobic cluster and interacts with W46G, L81G, C86G, I141G, L150G, and I202G, whereas D61R from the protease recognition sequence forms a specific interaction with the guanidinium group of R142G. Three further ionic interactions K42R-E148G, R55R-D82G, and R93R-D182G are likely to contribute to binding and enhancing specificity.
The interacting polar network between the catalytic triad, C45G, H151G, and D169G, is formed in the Rab32:GDP:GtgEC45A-complex once C45G is modeled to the active site cysteine nucleophile (Supplementary Fig. 11)6,7,8. Furthermore, the oxyanion hole can be unambiguously assigned to the peptide backbone NH-groups of C45G and N44G as well as the side chain Q33G Nε2. (Fig. 3f), while the thiol-group of C149G potentially acts as the Brønsted acid to generate a proper N-terminal leaving group of V60R (Fig. 3f).
In summary, GtgE interacts tightly with Rab32 via three binding platforms referred to as spike 1–3 that accommodate the switch regions. In particular, the resolved structure reveals that the cleavage site of Rab32 is pulled into a deep cavity formed by spike 1 and 2.
Confirmation of the interface of the Rab32:GDP:GtgE complex
The crystal structure of Rab32:GDP:GtgEC45A-complex suggests that many side chains within the switch region contribute to the interaction between the GTPase and the protease. In order to verify the significance of individual amino acids in complex formation, we investigated Rab32 point mutants for the ability to form a complex with GtgE by yeast-two-hybrid (Y2H) experiments using the full-length proteins (see Fig. 3e for positions of mutants). We first established an Y2H-based interaction assay by combining the proteolytically inactive mutant C45AG of GtgE as prey and the Rab32-subfamily together with Rab23 as bait (Fig. 4a). Complex formation is detected by growth on histidine-lacking media only for the C45AG mutant, but not the wild-type GtgE, indicating an interaction exclusively between the GTPase substrates and the proteolytically inactive GtgE. Consistent with our biophysical analyses we observe that this interaction is of high affinity for Rab32 since growth was also observed under more stringent conditions in the presence of 10 mM 3′-AT (Fig. 4a). The interaction with Rab38 is also detected, albeit the positive signal is more susceptible to the presence of 3′-AT, indicating a weaker interaction. Rab29 is self-activating in the Y2H assay (left panel) and thus cannot be used for monitoring GtgE-binding.
Next, a library of Rab32 switch-region mutants was explored by Y2H-based interaction studies in order to verify the interface observed in the complex crystal structure. In the switch I region, the amino acid mutants T52AR (P3), I58R (P2), G59R (P1), D61R (P2′), and F62R (P3′) show a strong growth defect, confirming that they are involved in forming the Rab32:GDP:GtgEC45A-complex (denominations in parentheses correspond to the Schechter-Berger-nomenclature for positions in the protease substrate cleavage site) (Fig. 4b, top). In particular, mutation of I58R, identified as a hub for hydrophobic interaction in the crystal structure, does not support growth in Y2H, thereby supporting the complex interface and the significance of this residue.
We also conducted a Rab32 switch II mutational analysis by employing the Y2H-screening (Fig. 4b, bottom). However, none of the mutants directly involved in the complex interface demonstrate a detectable growth defect, indicating that switch II mutations may be tolerated due to the high complex affinity. In contrast, the mutants W80AR and D81AR in the interswitch-region show severe growth inhibition. The side chain of W80R is a constituent of the important Rab hydrophobic triad that is relevant for binding to most interaction partners22. Since the W80R side chain interacts with F62R, the W80AR mediated growth defect can be explained by an influence on the positioning of the switch I F62R side chain that is important for GtgE binding (Fig. 4b, top).
In addition to the Y2H-assays, we probed the structure-activity relationship of the complex with mutants in all interacting regions (Rab switch-regions and GtgE-spikes, Fig. 3e) by studying GtgE-mediated Rab32-proteolysis (Fig. 4c, d; Supplementary Figs. 1, 13; Supplementary Table 3). All designed mutations in Rab32 were either constructed based on the structural analysis of the Rab32:GtgE complex (K43AR, D61AR, R93AR) or inspired by the alignment between Rab32 and the non-GtgE substrate Rab23 (A56KR, L64ER, R87ER) (Fig. 4c, Supplementary Fig. 1). The tested Rab32-mutants display at least a 10-fold activity reduction, supporting their significance for the substrate recognition and complex formation. Mutants D61AR and L64ER show a 30-fold decrease in activity with GtgE, emphasizing the importance of the salt bridge between D61R and R142G. The L64ER-mutation most likely induces a steric hindrance between F34G and L41G, which further give insights into the Rab-substrate restriction. The mutants K43AR and R93AR abrogate salt bridge formation with E148G and D182G, respectively, hence confirming the relevance of the corresponding ionic interaction for complex formation. In addition, the decrease in activity for the R87ER mutant supports the importance of the arginine side chain in binding to GtgE spike 3 via a backbone H-bridge with K194G. The mutant A56KR presumably causes steric clashes with the side chains of L150G and P143G that form the basis of the important spike 2 loop of GtgE (Fig. 3e), thereby corroborating the relevance of this region for recognition of Rab32 by the protease.
Moreover, all GtgE-mutants except for E148AG show a significant drop in activity (Fig. 4d). In particular, Q33AG decreases the activity 200-fold, confirming its key role in the oxyanion hole. The importance of the D61R-R142G salt bridge as a central interaction site mentioned before is once more confirmed using the R142AG mutation. The C149AG mutant reduces the activity by a factor of 2. Its localization near the active site suggests a role as a proton donor for the newly formed N-terminus during catalysis. The salt bridge formed by D182G with the switch II residue R93R shows a greater impact in the complex formation than the one constituted by E148G with K43R. The residue K186G is indirectly involved in the complex interface, stabilizing spike 3 by forming two intramolecular polar contacts with Q178G and E153G near the catalytic base H151G, rendering K186G an important organizing element in the switch II/spike 3-interface.
The Rab32:GDP:GtgE complex structure was also used to obtain insights into the Rab-selectivity. On the basis of the amino acid sequence alignment between the GtgE-substrates Rab32, 29, 38 and the non-GtgE-substrate Rab23, we identified four amino acids that are located in the protein–protein interface and may be incompatible with binding of Rab23 to the protease (Supplementary Fig. 1). Therefore, we tested whether Rab23 is converted into a GtgE-substrate by introducing the Rab32 analogous substitutions K40A, E48L, Q50V, and E70R in vitro. However, this Rab23 mutant was not cleaved by GtgE, indicating a multifactorial selection mode that cannot be reduced to a few specific amino acid side chains (Supplementary Fig. 14).
Thus, the mutational analysis confirms the Rab32–GtgE-interface observed in the complex crystal structure.
Insights into the Rab32:GDP specificity of GtgE
We next addressed the molecular basis for the strict preference of GtgE for the inactive GDP-bound Rab32 by analyzing the Rab32:GDP:GtgEC45A-complex crystal structure (Fig. 1b). Three potential hubs on Rab32 were identified that could hypothetically contribute to the targeting of GtgE to the inactive GTPase: hub 1 is constituted by the switch II residue E86R, hub 2 by the switch I amino acid Y54R, and hub 3 by the switch II side chain of F88R together with K194G and E114G (see below, Fig. 5a, b).
Close inspection of a structural superposition of the complex with active, GppCH2p-bound Rab32 suggests a potentially important role of the hub 1 side chain of E86R in activity-state discrimination by GtgE17. In the GtgE-bound form, the side chain of E86R protrudes into a Rab32-cavity that would be occupied by the γ-phosphate of GTP or GppCH2p in the active state (Fig. 5a). Consequently, the conformational rearrangement of switch II by GtgE-binding would result in a steric and electrostatic clash of E86R with the γ-phosphate, thereby making binding of Rab32:GTP to GtgE incompatible. In order to test this hypothesis, we analyzed the GtgE-mediated proteolysis of the E86AR-mutation of Rab32 with the full-length proteins. Surprisingly, the E86AR-mutation did not change the activity-state preference of GtgE since only GDP-bound but not GppNHp-loaded Rab32-E86AR was a substrate as monitored by SDS-PAGE based proteolysis (Fig. 5c). Hence, additional factors contribute to the Rab32:GDP specificity of GtgE.
As a further potential decisive element of the nucleotide discrimination, we focused on the switch I amino acid Y54R as hub 2. The Y54R-position is conserved among small GTPases and forms a polar interaction with the GTP γ-phosphate in the active state. In the Rab32:GDP:GtgEC45A structure, Y54R is turned away from the phosphates, pointing towards the guanine base. This observation could imply that Y54R is unable to support GTP-coordination in the Rab32:GtgE complex and therefore does not permit binding of GtgE to Rab32:GTP. Thus, the full-length mutants Y54FR and Y54AR were loaded with GDP or GppNHp and tested for their substrate properties towards GtgEWT (Fig. 5d and Supplementary Fig. 15A). Surprisingly, we find that again only the inactive GDP-bound Rab32-mutants were cleaved, and even after long incubation time no hydrolysis could be detected for the GppNHp-bound mutants, rendering the Y54R unimportant in the nucleotide-state selection.
In order to obtain additional insights into the nucleotide-state specificity of GtgE, we performed atomistic molecular dynamics (MD) simulations. We first considered calculating the structure of a hypothetical complex between active Rab32:GTP and GtgE. Therefore, the only currently available structure of active Rab32 bound to the non-hydrolysable GTP-analog GppCH2p was taken from its complex with the effector protein VARP (PDB ID: 4CYM)17. Replacement of the structure of the GppCH2p-bound Rab32 from its complex with VARP resulted in steric clashes of Rab32 with GtgE despite the global structure of Rab32 closely resemble each other in the two X-ray structures (Supplementary Fig. 16). Instead of probing the formation of Rab32:GTP:GtgE complex by extensive free energy calculations, we computationally replaced GDP with GTP in the nucleotide binding pocket of Rab32 and probed the structural determinants for GtgE-dissociation from the complex. This approach is justified since small GTPases can form complexes with their binding partners in both activity states albeit with differing affinities23. It is common, however, that protein–protein affinities are mainly determined by the rate of complex dissociation (k off, 10−5–103 M−1 s−1) rather than by the second-order rate constant of complex association (k on, 106–107 M−1 s−1). The MD-simulation, therefore, mimics the initial molecular steps of GtgE dissociating from Rab32:GTP, and aims to probe determinants that would contribute most to dissociating the Rab32:GTP:GtgE complex once formed. Notably, these computations do not address the rate of complex association, since they would require extensive free energy simulations, and are outside the scope of the present study.
A principal component analysis based on the MD-trajectories of the Rab32:GtgE complex confirms that the Rab32:GtgE complex is preserved in the GDP-state. However, replacement of the nucleotide for GTP alters the complex behavior and reveals loop-motions of Rab32 and GtgE into opposite directions (Supplementary Fig. 17). Consequently, the analysis suggests that Rab32 is binding to GtgE in the GDP-state but dissociates from the protease in the GTP-state.
In our ca. 200 ns MD-simulations, the global structure and dynamics of both Rab32 and GtgE remain similar, but we observe an increase in distance between F88R and K194G in the GTP-bound complex (Fig. 5e, bottom and Supplementary Fig. 18A). The MD-simulations reveal that a third putative hub (hub 3) in the complex interface between switch II and spike 3 could be involved in the Rab32–GtgE recognition, comprising residues F88R, K194G, and E114G (Fig. 5b). In the MD-simulations with GTP, we observe an increase in the F88R–K194G distance, which could lead to a weaker Rab32–GtgE interaction, consistent with the observed lack of Rab32:GTP-binding to GtgE in the in vitro experiments (Supplementary Fig. 18B). In order to probe the function of F88R in Rab32, we replaced this residue with a glycine (F88GR), and re-initiated the MD-simulations. Interestingly, the MD-simulation of the F88GR demonstrate that the GTP- and GDP-states behaves more similar in this mutant (Supplementary Fig. 18C). Therefore, to validate the significance of hub 3 for the GtgE selectivity towards GDP over GTP, we produced the Rab32 F88GR-mutant. To this end, we purified the full-length Rab32 wild type and the F88GR-mutant with GDP or GppNHp, challenged it with cleared E. coli cell lysates containing overexpressed full-length wild-type or GtgE-mutants (E114AG, K194AG, E114AG/K194AG), and analyzed Rab32-proteolysis using SDS-PAGE based gel shift (Fig. 5f and Supplementary Fig. 15B). Interestingly, although Rab32 F88GR generally becomes a worse substrate for GtgE, the F88GR mutation indeed results in a loss of the strong GDP-state preference of GtgE for Rab32 (Fig. 5f). The decrease in GtgE-efficiency for Rab32 F88GR is expected due to the involvement of F88R in the complex interface. However, the F88GR-mutant is now cleaved both in the GDP- and GTP-states in contrast to the wild-type GTPase (Fig. 5f). Thus, the F88R-moiety is obviously important in the switch II rearrangement during complex formation, as indicated by the MD-data.
In contrast, the alanine substitutions of GtgE E114G and R194G involved in hub 3 show opposite effects and are neutralizing each other in the double mutant when combined with Rab32 F88GR (Supplementary Fig. 15B, right). The increased space in the F88GR mutant may be required to form a catalytically competent Rab32:GTP:GtgE complex. We speculate that E114AG decreases the proteolytic activity emphasizing its vital role as electrostatic anchor for K194G. When not fixed by the carboxylate function of E114G, the ε-amino group of K194G is free to establish other intramolecular salt bridges or polar contacts with the neighboring D193G or Y195G, and could thereby disturb the organization of the hub 3 interface. The Rab32 switch II interaction with spike 3 may thus be a crucial part in the nucleotide selection.
In conclusion, we provide insights into a multifactorial selection mechanism since no single tested hub was able to abolish the nucleotide substrate preference of GtgE entirely. Nevertheless, the computationally derived and experimentally verified F88GR-mutation results in the proteolysis of both active and inactive Rab32.
In this work, we have biochemically and structurally characterized the Rab32:GDP:GtgE complex. We have established that GtgE specifically acts on inactive, GDP-bound Rab32. Furthermore, based on combined atomistic molecular simulations and in vitro experiments, we could demonstrate that the F88R position in the switch II regulatory region of Rab32 contributes to a multifactorial activity-state discrimination mechanism by GtgE and that it forms a basis for the GDP-state specificity of the protease.
Only two other proteins are known to have exclusive affinity for inactive-state Rabs. Here, the functionally and structurally related regulatory factors Rab escort protein (REP) and GDI are involved in the prenylation and membrane delivery or membrane cycling of Rabs, respectively, and interact with low nanomolar affinity with Rab:GDP24,25,26. In particular, the physiological role of GDI is to effectively solubilize lipidated Rabs from intracellular membranes after the protein has returned to the inactive form. Even though the complexes of Rab-proteins with REP and GDI have been extensively investigated on a structural level, the molecular basis for the exclusive interaction with inactive Rabs remains elusive20,27,28. Our structural and molecular dynamics analysis of the Rab32–GtgE interactions, however, revealed that the side chain of F88R in Rab32 forms an important structural determinant for discriminating the active and inactive Rab GTPase.
It is interesting to note that the Rab-recruitment to intracellular membranes is dependent on membrane-localized GEFs. Therefore, Rab-targeting and activation are directly coupled29. Consequently, GEF-mediated Rab32 recruitment to the SCV would result in the creation of active Rab32:GTP that is not a GtgE-substrate and consequently would not be subject to proteolysis. Interestingly, Salmonella secretes another protein that ensures inactivation of Rab32:GTP, making Rab32:GDP accessible to GtgE. During infection, the Salmonella Rab32-GAP SopD2 is released and catalyzes the conversion of Rab32:GTP to Rab32:GDP15. Apparently, Salmonella causes effective GtgE-mediated Rab32-cleavage by ensuring the prior conversion of active into inactive Rab32.
The question arises as to why Salmonella has evolved with a GDP-state specific Rab-protease making it dependent on a second secreted protein (i.e. SopD2). We hypothesize, however, that the presence of SopD2 ensures time-efficient access to GDP-loaded Rab32 since the latter would be blocked by tightly binding to GDI. We therefore suggest the following order of events: The Rab-protein is targeted to and activated at the membrane by a corresponding GEF29. This successive pathway of GEF-mediated Rab32 activation/recruitment and SopD2-mediated Rab32-inactivation results in direct access of GtgE to Rab32 (Fig. 6). In the hypothetical case where GtgE was to attack Rab32:GDP without prior action of GEFs and SopD2, it would need to directly compete with GDI for Rab32-binding in the cytosol. Since GDI and GtgE occupy the identical binding site on Rab32 and since the interaction of Rabs with GDI are of high nanomolar affinity, GDI would effectively shield Rab32 from GtgE20,26,27. The high affinity of the Rab32:GDP–GtgE interaction (indicated by a nanomolar K D, K D, Rab32:GDP:GtgE = 100 nM) would not be sufficient to displace GDI (K D,Rab:GDI < 5 nM) in the cytosol and would possibly not allow time-efficient Rab32-cleavage.
In contrast, the combination of SopD2 and GtgE Salmonella at the SCV membrane during infection guarantees fast Rab32-inactivation as a results of a series of rapid enzymatic steps (Fig. 6): Rab32:GDP is likely recruited to the SCV via a GEF (BLOC-3) from its cytosolic complex with GDI promptly. The resulting pool of active Rab32:GTP is then immediately targeted by the GAP SopD2, leading to GTP-hydrolysis and the production of membrane-localized Rab32:GDP. Finally, GtgE is attacking Rab32:GDP at the SCV and the proteolyzed GTPase is formed quickly. Therefore, the cooperation of SopD2 and GtgE may overall be the most efficient way for quantitative Rab32-proteolysis (Fig. 6). An efficient Rab-inactivation therefore appears to be crucial for the progressing SCV maturation.
Together with previous findings, our present results contribute to the understanding of the order of molecular events at the SCV: the maturation of the SCV is dependent on the small GTPases Rab7 and Rab930,31,32. After activation, Rab9 may recruit the Rab32-GEF BLOC-3 to the surface of the SCV33,34. BLOC-3 subsequently recruits Rab32 to the SCV from its complex with GDI and activates the G-protein by loading with GTP. In order to stably establish the infection, Salmonella counteracts Rab32-dependent membrane pathways with SopD2 and GtgE, leading to Rab32 conversion into the GDP-state and irreversible proteolysis in the switch I region, respectively5,15.
It is interesting to note that despite of Rab32 recruitment to the SCV, the protein is not stably accumulated at the membrane in Salmonella Typhimurium (containing the GtgE and SopD2 proteins), indicating that GtgE-cleaved Rab32:GDP is lost from the vacuole in an active or passive process5. Likely, the Rab32-proteolysis is not affecting intrinsic membrane affinity since the C-terminal lipids remain unaffected, making a direct GtgE-mediated Rab membrane release improbable. Since our results demonstrate that cleaved Rab32 remains a stable protein that alsocan be purified in vitro, active processes to remove the Rab-protein from membrane may be ongoing. Here, ubiquitination in conjunction with proteasomal degradation or membrane extraction by GDI are conceivable mechanisms. Cleaved Rab32 may still interact with and be extracted by GDI, leading to effective removal of the GTPase. GDI is only weakly interacting with the switch I regions of Rab-proteins and therefore presumably tolerating the proteolytic modification20.
We speculate that once extracted by GDI, the GtgE-cleaved Rab32 is presumably not recruited by BLOC-3 to the SCV again. Although the 3D-structure of BLOC-3 is unknown, its sequence similarity to components of the Ypt7–GEF-complex Mon1/Ccz1 suggests that the interaction with Rab32 happens in a mechanistically identical manner35. In contrast to the Rab–GDI interactions, the switch I loop intimately interacts with the GEF. Consequently, Rab32-proteolysis is potentially incompatible with activation by BLOC-3, and thus re-recruitment of GDI-bound Rab32 would be prohibited.
In conclusion, GtgE is a GDP-state specific Rab32-protease having a high affinity for its substrate. Structural studies and MD-simulations identify an amino acid hub that contributes to the selective binding of the inactive state of Rab32. The analysis provides insights into the order of protein interactions involved at the SCV and imply a requirement for the Salmonella GAP SopD2 in order to permit GtgE-mediated Rab32-proteolysis. It will be interesting to investigate how the GtgE-processed Rab32 contributes further to the infection process.
For all cloning tasks, either XL1 blue or Mach1 E. coli strains were used. Site and ligation independent cloning (SLIC) approach was applied for all molecular cloning. The gtgE gene from Salmonella enterica serovar Typhimurium str. SL1344 (CBW17091.1) was synthesized with an optimized sequence for E. coli codon usage provided in a standard cloning vector (Life Technologies). The sopD2 gene (WP_001145561.1.) was amplified from genomic DNA from the same strain. Both Salmonella genes were subcloned for heterologous expression into a modified pMAL-c2X vector (Addgene ID 75286) harboring an N-terminal His6-MBP tag (MBP: maltose binding protein) fused through a TEV protease cleavage site between the 3′-NdeI and 5′-HindIII restriction sites. Additionally, truncated Rab3218-201 and GtgE21-214,C45A were generated in the pMAL vector as described above and used for protein crystallization and structural studies, only. For Yeast-Two-Hybrid experiments, gtgE was introduced to the pGAD-HA vector (Clontech, Cat. No 630442) harboring an N-terminal activation domain. The human Rab GTPases (Rab32, NP_006825.1; Rab29 (also Rab7L1) NP_001129134.1; Rab38 NP_071732.1; and Rab23 NP_057361.3) were introduced into the pMAL vector with 3′-NdeI and 5′-XhoI restriction sites. Fusion constructs of the pLexA vector (Addgene ID 11342) with all mentioned GTPases harbored an N-terminal DNA binding domain lacking the C-terminal prenylation site cysteines. Mutations were introduced with the Q5® Site-Directed Mutagenesis Kit (NEB) following the manufacturer’s procedure. All plasmids were verified by DNA sequencing (GATC Biotech).
Recombinant protein expression and purification
All proteins were recombinantly expressed in E. coli BL21(DE3)-RIL chemically competent cells (Agilent) growing in LB broth supplemented with ampicillin. The expression culture was inoculated with 2–5% (v/v) from a pre-culture and grown at 37 °C shaking at 180 r.p.m.. Expression was induced at an OD600 of 0.5 by addition of 0.3 mM IPTG (final). After induction the culture rested at 4 °C for 30 min and was grown 16–20 h at 22 °C and collected at 5900 × g for 25 min. The pellet was washed with phosphate-buffered saline (1xPBS) and collected again under same conditions. Cell pellets were snap frozen in liquid nitrogen and stored at −20 °C until further use. Pellets of GtgE and SopD2 proteins were resuspended in 10 mL Buffer A (50 mM Hepes pH 7.5, 500 mM NaCl, 2 mM β-mercaptoethanol (BME)) for each gram cells and lysed mechanically with one passage through a fluidizer at 2.0 kbar (Constant systems). Lysate was cleared by centrifugation 48,000 × g for 30 min and the supernatant was applied to a 5 mL Ni2+-immobilized metal-affinity chromatography column (IMAC) column (HiTrap, GE Healthcare) previously equilibrated with Buffer A. The column was washed with 6 column volumes (CV) 5% Buffer B (Buffer A supplemented with 500 mM imidazole) and protein of interest was eluted applying a gradient up to 50% Buffer B in 10 CV. Fractions containing the His-MBP tagged protein of interest were pooled and dialyzed over night at 4 °C against Buffer C (20 mM Hepes pH 7.5, 50 mM NaCl, 2 mM BME) and subjected to Tobacco Etch Virus (TEV)-protease cleavage. Cleaved protein mixture was applied a second time on an IMAC column to remove the TEV protease and His-MBP tag. The highly pure protein was collected in the flow through, concentrated, snap frozen and stored at −80 °C.
Purification of Rab GTPases: The procedure was the same as described for the GtgE purification above except all buffers were supplemented with 1 mM MgCl2 and 10 µM GDP and after cell lysis 1 mM phenylmethylsulfonyl fluoride (PMSF; final) was added before centrifugation. To reach homogeneity after the IMAC Rab GTPases were further subjected to size-exclusion chromatography (SEC) on a Superdex 75 (16/600) (GE Healthcare) pre-equilibrated with SEC-buffer (20 mM HEPES, pH 7.5, 50 mM NaCl, 1 mM MgCl2,10 µM GDP, and 1 mM DTT). Fractions containing monomer were concentrated, snap frozen and stored at −80 °C. The identity of all proteins was verified by mass spectrometry using an electron-spray-ionization mass spectrometer coupled to a liquid chromatography (LC-ESI-MS; LCQ fleet; Thermo Scientific).
For the in vitro complex formation, 8 µM Rab GTPase was incubated with 8 µM GtgEC45A in SEC-buffer for 30 min at room temperature. For analysis 50 µL of the mixture was subjected to an analytical gel filtration (Superdex 75 13/30, GE Healthcare) pre-equilibrated with SEC-buffer with a flow of 0.5 mL/min recording the absorption at 280 nm. As negative controls, the respective monomers were analyzed identically.
Yeast-Two-hybrid interaction studies
For the Y2H assay, pLexA-GTPase plasmids (bait) encoding for full-length Rab proteins were transformed with MAT-a yeast strain Y187 adopted from a previous method36 and plated on selection dropout medium lacking tryptophan (SD-W)37. In brief, 700 µL of stationary yeast culture was collected at 375 × g for 2 min and vortexed with 100 µL sterile one-step buffer (0.2 M lithium acetate, 40% PEG 3350, and 100 mM DTT). Subsequently 100–500 ng of respective plasmid DNA was added and incubated at room temperature for 15 min before heat shock was applied for 30 min at 45 °C. The pGAD-gtgEC45A plasmids (prey) encoding for inactive full-length GtgE protein were transformed the same way with MAT-α S. cerevisiae L40∆Gal4 and were grown on SD-L plates (lacking leucine) at 30 °C for 3–4 days. For yeast mating, single colonies of a prey and a bait clone were grown separately in appropriate selection medium for 2 days while shaking. To 100 µL YAPD media37 in a 96-well plate (Sarstedt) 50 µL of bait and prey culture were added sequentially and incubated 22–24 h at 30 °C with 180 r.p.m. Cells were resuspended and 5 µL were spotted on SD-LW plates as mating control and SD-LWH (lacking histidine) for phenotypic read out. Plates were analyzed after 3–4 days of incubation at 30 °C. 3-Aminotriazol (3′-AT, Sigma Aldrich), an inhibitor of the histidine biosynthesis, was added to SD-LWH agar plates in the range of 0–10 mM probing the interaction intensity qualitatively.
Isothermal titration calorimetry
Interaction studies by isothermal titration calorimetry (ITC) were conducted on a ITC200 microcalorimeter (MicroCal). Measurements were performed in SEC-buffer at 37 °C. Rab32FL:GDP was diluted to 20 µM and GtgEFL,C45A was introduced to the injection unite with 200 µM. Injection volume was set to 1.5 µL per injection and the heat power was recorded over time until binding saturation was obtained. The binding isotherms were integrated, corrected for the offset, and the data were fitted to a one-site-binding model using the ITC analysis software provided by the manufacturer (MicroCal), yielding the equilibrium-dissociation constant (K D).
Interaction studies by analytical ultracentrifugation (AUC) were conducted in a Beckman ProteomeLab XL-A analytical ultracentrifuge (Beckman Coulter) equipped with a fluorescence detection system (Aviv AU-FDS; Aviv Biomedical) as described previously38. A constant concentration of 200 nM Rab32FL-Atto488 in SEC-buffer was incubated with varying molar ratios of GtgEFL,C45A as indicated for 2 h in the dark at room temperature, ensuring equilibrium conditions prior to measurement. The data were analyzed with SedFit software39. The sedimentation coefficients and molecular weights of the present species were obtained by solving the Lamm equation. The fraction of labeled Rab-proteins bound to GtgEC45A was plotted against the GtgEC45A concentration and data were fitted with a hyperbolic function for calculation of the K D.
Nucleotide loading of GTPases
Nucleotide exchange of small GTPases with GDP or GTP (Carbosynth) was performed in SEC-buffer supplemented with 10 mM EDTA and a 40-fold molar excess of nucleotide over protein and incubation at 15 °C for 5–16 h before performing size-exclusion chromatography (Superdex 75 16/60; GE Healthcare). Nucleotide exchange with the non-hydrolyzable analog GppNHp (Jena Bioscience) was performed by addition of equimolar amounts to the GTPase with 5 U Antarctic Phosphatase (NEB) per mg protein in nucleotide exchange buffer (50 mM Tris HCl pH 8.0; 200 mM (NH4)2SO4; 10 µM ZnCl2). Accidentally precipitated protein was removed by centrifugation and a buffer exchange was performed on a NAP 5 desalting column (GE Healthcare) according to the manufacturer’s instructions with SEC-buffer containing 1 µM GppNHp. Protein containing fractions were pooled, concentrated, snap frozen, and stored at −80 °C. Nucleotide loading efficiency was tested by ion-pairing reversed-phase high performance chromatography (RP-HPLC; Prontosil C18, Bischhoff Chromatography) in 50 mM potassium phosphate buffer pH 6.6; 10 mM tetrabutylammonium bromide and 12% acetonitrile (v/v). Protein samples (25 µM, 30 µL) were heat precipitated at 95 °C for 5 min and cleared by centrifugation for 10 min at 48,000 × g. Supernatant was subjected to chromatographic separation and nucleotide peaks were integrated and normalized to the total amount of nucleotides detected set to 100%. Nucleotide retention times were determined with the respective nucleotide standard in a separate run.
Kinetics of Rab32 cleavage
Cleavage of Rab GTPases by GtgE was monitored by SDS-PAGE gel shift assay. Typically 8 µM of RabFL was cleaved by 8 nM GtgEFL in SEC-buffer supplemented with 10 µM GDP, 10 µM GTP or 1 µM GppNHp respectively at 25 °C. At different time points, 10 µL samples (ca. 2 µg) were taken an the reaction was quenched immediately with pre-heated Laemmli buffer and 3 min boiling at 95 °C. For quantification purpose 2 µM, MBP was added as internal standard into the reaction mixture prior to start of the experiment. After Coomassie staining (Brilliant blue R-250, Thermo Scientific), band intensities were determined by densitometry from gray values with image studio lite software (LI-COR). Rab-bands were normalized with the respective MBP-band intensity and plotted over time. The apparent rate constant (k obs) was obtained by fitting the data exponentially and k cat/K M resulted from dividing k obs by GtgE concentration [E 0]. For rapid qualitative analysis of GtgE-mutant activity, the method was adopted as follows: GtgE variants were recombinantly expressed in 50 mL. Cells equal to 5 OD units were resuspended in 0.5 mL SEC-buffer lacking nucleotides and lysed by ultrasonication (Digital Sonifier® S-450, Branson Ultrasonics). Protein expression levels were monitored from the cleared lysate by SDS-PAGE and total protein amounts were determined with the OD at 280 nm. Subsequently, the lysate was introduced directly to the assay as described above (ca. 2 µg in 10 µL reaction) or diluted 1:100 in nucleotide free SEC-buffer prior to use.
GtgE-mediated Rab32 cleavage
The applicability of fluorescence-based methods for monitoring GtgE-mediated Rab32-proteolysis was evaluated using a potential change in intrinsic Rab32-tryptophane fluorescence or a change of mant-fluorescence for Rab32 loaded preparatively with the fluorescent GDP-derivative mantGDP.
Cleavage experiments of Rab32:GDP by GtgE were performed at 25 °C in a 1 mL quartz cuvette using a fluorescence spectroscope (FluoroMax®-4, Horiba Ltd., Japan). All measurements were performed in filtered degassed assay buffer (20 mM HEPES, pH 7.5; 50 mM NaCl; 1 mM MgCl2; 1 mM DTT) with nucleotides added right before the measurement. For tryptophane fluorescence experiments, 500 nM Rab32:GDP was added to the assay buffer supplemented with 10 µM GDP. Samples were excited at 297 nm (1 nm band-pass) and emission was recorded at 340 nm (5 nm band-pass). For mant-fluorescence experiments, 500 nM Rab32:mGDP was added to the assay buffer supplemented with 20 nM mGDP. Samples were excited at 340 nm (1 nm band-pass) and emission was recorded at 440 nm (5 nm band-pass). GtgE was added at 10 nM as final concentration to start the reaction (GtgE:Rab molar ratio 1:50).
Covalent protein fluorescence labeling
GtgEFL,C45A was buffer exchanged to labeling buffer (20 mM HEPES pH 7.5, 50 mM NaCl, 1 mM MgCl2; 1 mM Tris(2-carboxyethyl)phosphine (TCEP)) and subsequently labeled covalently with the primary lysine amines using NT-647-NHS labeling kit (Nanotemper) following the manufacturer’s instructions. The reaction was quenched by addition of 10 mM Tris HCl pH 7.5 and buffer exchanged on a desalting column for labeling buffer. Rab32FL was exchanged to labeling buffer supplemented with 10 µM GDP. Thiol groups in Rab32FL were covalently labeled with 1.1 molar excess of Atto488-Maleimide (ATTO-TEC) incubating at 4 °C for 16 h and the reaction was quenched by addition of 0.5 mM DTT.
Lysine methylation of the Rab32:GDP:GtgE complex
Methodology was adopted from a previous publication16. The lysine methylation reaction was performed in SEC-buffer with the preformed full-length Rab32:GDP:GtgEC45A and Rab3218-201:GDP:GtgE21-214,C45A hetero complexes (1:1 molar ratio) and 2 mg mL−1 total concentration. 20 µL freshly prepared 1 M dimethylamine-borane complex in reaction buffer (5 µL final; ABC; Sigma Aldrich) and 40 µl 1 M formaldehyde (8 µL final; made from 37% stock in reaction buffer; Sigma) were added per milliliter protein solution, and the reaction was gently rotated at 4 °C for 2 h. A second addition of ABC and formaldehyde followed a 2 h incubation under same conditions. Following a final addition of ABC the reaction was rotated overnight at 4 °C. Precipitations were removed by centrifugation and the supernatant was applied to a gel filtration (Superdex 75 16/60, GE Healthcare) pre-equilibrated with SEC-buffer. Fractions containing the hetero-complex were concentrated to a total of 10 mg/mL and immediately applied to protein crystallization.
Differential scanning fluorimetry
For the protein stability measurements, differential scanning fluorimetry (DSF) was conducted on a real time PCR device (Mx3005P, Agilent). In each well of technical replicas, 2 µg of Rab32FL:GDP (full length or cleaved) was prepared in 20 µL SEC-buffer supplemented with 1-fold SYPRO Orange (Sigma Aldrich, 5000-fold stock). The mixture was heated in 1 °C min−1 steps from 25–95 °C while recording the fluorescence signal with the excitation at 465 nm and the emission at 590 nm. The protein melting temperature (T m) was obtained by fitting the data with the Boltzmann fit.
Structure determination by X-ray crystallography
Crystallization trials of full-length and truncated human Rab32:GDP (residues 18–201 in the truncated construct) complexed with the full-length and truncated protease GtgEC45A mutant from Salmonella enterica ssp. Typhimurium (residues 21–214 in the truncated complex) were performed by the sitting-drop vapor diffusion method, using protein concentrations of 10 mg/mL. Initial sparse–matrix screens with 0.2 μL of protein and 0.2 μL of reservoir solution identified several hits, but these crystals did not diffract X-rays beyond 5 Å resolution. Reductive methylation on surface exposed lysine residues significantly improved the diffraction quality of the full length and truncated Rab32:GDP:GtgE-complexes16. Suitable crystals of the full-length complex were grown in 0.05 M imidazole pH 8, 20% PEG 6000; crystals of the truncated complex were obtained from 0.1 M Bis-Tris pH 5.5, 25% PEG 3350. Crystals of both complexes were cryoprotected in reservoir solution supplemented with 25% ethylene glycol and flash cooled in a stream of liquid nitrogen gas at 100 K.
Diffraction data collection and structure determination
The data sets of full length and truncated Rab32:GDP:GtgE complexes were collected from single crystals using synchrotron radiation (λ = 1.0 Å) at the X06SA-beamline (Swiss Light Source, Villingen, Switzerland). X-ray intensities were assessed and data reduction carried out with XDS and XSCALE40. Crystals of the full-length complex diffracted to a resolution of 2.9 Å (space group P6522 with a = b = 67.4 Å, c = 427.3 Å); crystals of the truncated complex diffracted to a resolution of 2.3 Å (space group P212121 with a = 47.8 Å, b = 66.9 Å, c = 110.9 Å).
Initial phases were calculated for the Rab3218–201:GDP:GtgE21–214,C45A complex by Patterson search procedure using the deposited coordinates of Rab32:GppCH2p (residues 18–201; PDB ID: 4CYM) and a partial model of GtgE (residues 80–213; PDB ID: 4MI7). The starting models were sequentially placed as a heterodimer into the asymmetric unit using PHASER (TFZ score = 11.7)41. After model building with COOT42, rigid body refinement and preliminary positional refinement with phenix.refine43, the GtgE subunit was re-traced. Application of feature-enhanced maps using PHENIX in the early and intermediate stages of model building and refinement aided greatly in extending the model to finally include residues 23–213 of GtgE and residues 22–195 of Rab32, along with the GDP and magnesium ion co-factors of the latter protein44. Model refinement was performed in phenix.refine and REFMAC45. Eventually, water molecules were placed automatically using ARP/wARP46. The final model of the Rab3218–201:GDP:GtgE21–214,C45A complex, with refined Translation/Libration/Screw (TLS) tensors and individual B-factors, converged to R work and R free values of 20.8% and 24.5%, respectively. Coordinate geometries were confirmed via MOLPROBITY to possess good stereochemistry and small bond-length and angle RMSDs, with a single glycine residue 59 of Rab32—adjacent to the site of proteolytic cleavage by GtgE—in the disallowed region of the Ramachandran plot47. Magnesium ion coordination was verified with the CheckMyMetal server48.
The full-length Rab32:GDP:GtgEC45A structure was solved by molecular replacement, applying the refined coordinates of the Rab3218–201:GDP:GtgE21–214,C45A heterodimer as the initial search model. A single copy of the complex constituted the asymmetric unit of the crystals belonging to space group P65 22. In this case, model building was straight forward, eventually featuring residues 23–217 of GtgE and residues 21–195 of Rab32. TLS refinement yielded converging values for R work = 20.1% and R free = 24.0% along with low bond-length and angle RMSDs. Inspection of the Ramachandran plot showed good stereochemistry, with glycine residue 59 of Rab32 again the sole outlier. Structural representations of proteins were prepared with PyMol49 or Visual Molecular Dynamics (VMD) software50.
Molecular dynamics simulations
MD-simulations were performed based on our resolved X-ray structure of the truncated Rab32:GDP:GtgEC45A-complex (i.e., Rab3218–201:GDP:GtgE21–214,C45A) by replacing Ala-45 with Cys in order to model the wild-type GtgE. The protein complex was embedded in a water-ion environment with a 150 mM NaCl concentration, and the complete simulation setup comprised 98,980 atoms. The wild type and F88G truncated Rab32:GtgE complexes were simulated in presence of both GTP and GDP for 190 ns at T = 310 K using a 1 fs time step, by using the CHARMM36 force field51, and treating the long-range electrostatics using the Particle Mesh Ewald (PME) approach52. The MD-simulations were performed using NAMD53, trajectories were analyzed using VMD50, and principal component analysis based on the MD-trajectories were performed using ProDy54,55.
Structure factors and model coordinates for the full length and truncated Rab32:GDP:GtgEC45A-complexes have been deposited in the RCSB Protein Data Bank under the accession codes 5OED and 5OEC, respectively. The data that support the findings of this study are available from the corresponding author upon request.
Cherfils, J. & Zeghouf, M. Regulation of small GTPases by GEFs, GAPs, and GDIs. Physiol. Rev. 93, 269–309 (2013).
Aktories, K. Bacterial protein toxins that modify host regulatory GTPases. Nat. Rev. Microbiol. 9, 487–498 (2011).
Grassl, G. A. & Finlay, B. B. Pathogenesis of enteric Salmonella infections. Curr. Opin. Gastroen. 24, 22–26 (2008).
Ohl, M. E. & Miller, S. I. Salmonella: A model for bacterial pathogenesis. Annu. Rev. Med. 52, 259–274 (2001).
Spano, S. & Galan, J. E. A Rab32-dependent pathway contributes to Salmonella Typhi host restriction. Science 338, 960–963 (2012).
Spano, S., Liu, X. Y. & Galan, J. E. Proteolytic targeting of Rab29 by an effector protein distinguishes the intracellular compartments of human-adapted and broad-host Salmonella. Proc. Natl. Acad. Sci. USA 108, 18418–18423 (2011).
Kohler, A. C., Spano, S., Galan, J. E. & Stebbins, C. E. Structural and enzymatic characterization of a host-specificity determinant from Salmonella. Acta Crystallogr. D. 70, 384–391 (2014).
Xu, C., Kozlov, G., Wong, K., Gehring, K. & Cygler, M. Crystal structure of the Salmonella Typhimurium effector GtgE. PloS ONE 11, e0166643 (2016).
Raposo, G. & Marks, M. S. Melanosomes-dark organelles enlighten endosomal membrane transport. Nat. Rev. Mol. Cell Biol. 8, 786–797 (2007).
Wasmeier, C. et al. Rab38 and Rab32 control post-Golgi trafficking of melanogenic enzymes. J. Cell. Biol. 175, 271–281 (2006).
Raposo, G., Tenza, D., Murphy, D. M., Berson, J. F. & Marks, M. S. Distinct protein sorting and localization to premelanosomes, melanosomes, and lysosomes in pigmented melanocytic cells. J. Cell. Biol. 152, 809–823 (2001).
Bultema, J. J., Ambrosio, A. L., Burek, C. L. & Di Pietro, S. M. BLOC-2, AP-3, and AP-1 proteins function in concert with Rab38 and Rab32 proteins to mediate protein trafficking to lysosome-related organelles. J. Biol. Chem. 287, 19550–19563 (2012).
Holt, O. J., Gallo, F. & Griffiths, G. M. Regulating secretory lysosomes. J. Biochem. 140, 7–12 (2006).
Benado, A., Nasagi-Atiya, Y. & Sagi-Eisenberg, R. Protein trafficking in immune cells. Immunobiology 214, 507–525 (2009).
Spano, S., Gao, X., Hannemann, S., Lara-Tejero, M. & Galan, J. E. A Bacterial pathogen targets a host Rab-family GTPase defense pathway with a GAP. Cell Host Microbe 19, 216–226 (2016).
Walter, T. S. et al. Lysine methylation as a routine rescue strategy for protein crystallization. Structure 14, 1617–1622 (2006).
Hesketh, G. G. et al. VARP Is recruited on to endosomes by direct interaction with retromer, where together they function in export to the cell surface. Dev. Cell. 29, 591–606 (2014).
Meng, E. C., Pettersen, E. F., Couch, G. S., Huang, C. C. & Ferrin, T. E. Tools for integrated sequence-structure analysis with UCSF Chimera. BMC Bioinforma. 7, 339 (2006).
Pylypenko, O. et al. Structure of doubly prenylated Ypt1:GDI complex and the mechanism of GDI-mediated Rab recycling. EMBO J. 25, 13–23 (2006).
Rak, A. et al. Structure of Rab GDP-dissociation inhibitor in complex with prenylated Ypt1 GTPase. Science 302, 646–650 (2003).
Krissinel, E. & Henrick, K. Inference of macromolecular assemblies from crystalline state. J. Mol. Biol. 372, 774–797 (2007).
Merithew, E. et al. Structural plasticity of an invariant hydrophobic triad in the switch regions of Rab GTPases is a determinant of effector recognition. J. Biol. Chem. 276, 13982–13988 (2001).
Wu, Y. W. et al. Membrane targeting mechanism of Rab GTPases elucidated by semisynthetic protein probes. Nat. Chem. Biol. 6, 534–540 (2010).
Araki, S., Kikuchi, A., Hata, Y., Isomura, M. & Takai, Y. Regulation of reversible binding of smg p25A, a ras p21-like GTP-binding protein, to synaptic plasma membranes and vesicles by its specific regulatory protein, GDP dissociation inhibitor. J. Biol. Chem. 265, 13007–13015 (1990).
Regazzi, R., Kikuchi, A., Takai, Y. & Wollheim, C. B. The small GTP-binding proteins in the cytosol of insulin-secreting cells are complexed to GDP dissociation inhibitor proteins. J. Biol. Chem. 267, 17512–17519 (1992).
Wu, Y. W., Tan, K. T., Waldmann, H., Goody, R. S. & Alexandrov, K. Interaction analysis of prenylated Rab GTPase with Rab escort protein and GDP dissociation inhibitor explains the need for both regulators. Proc. Natl. Acad. Sci. USA 104, 12294–12299 (2007).
Ignatev, A., Kravchenko, S., Rak, A., Goody, R. S. & Pylypenko, O. A structural model of the GDP dissociation inhibitor rab membrane extraction mechanism. J. Biol. Chem. 283, 18377–18384 (2008).
Pylypenko, O. et al. Structure of Rab escort protein-1 in complex with Rab geranylgeranyltransferase. Mol. Cell. 11, 483–494 (2003).
Blümer, J. et al. RabGEFs are a major determinant for specific Rab membrane targeting. J. Cell. Biol. 200, 287–300 (2013).
Brumell, J. H. & Scidmore, M. A. Manipulation of Rab GTPase function by intracellular bacterial pathogens. Microbiol. Mol. Biol. Rev. 71, 636–652 (2007).
Meresse S., Steele-Mortimer O., Finlay B. B., Gorvel J. P. The rab7 GTPase controls the maturation of Salmonella typhimurium-containing vacuoles in HeLa cells. EMBO J. 18, 4394-4403 (1999).
Smith, A. C. et al. A network of Rab GTPases controls phagosome maturation and is modulated by Salmonella enterica serovar Typhimurium. J. Cell. Biol. 176, 263–268 (2007).
Gerondopoulos, A., Langemeyer, L., Liang, J. R., Linford, A. & Barr, F. A. BLOC-3 mutated in Hermansky-Pudlak syndrome is a Rab32/38 guanine nucleotide exchange factor. Curr. Biol. 22, 2135–2139 (2012).
Kloer, D. P. et al. Assembly of the biogenesis of lysosome-related organelles complex-3 (BLOC-3) and its interaction with Rab9. J. Biol. Chem. 285, 7794–7804 (2010).
Kiontke, S. et al. Architecture and mechanism of the late endosomal Rab7-like Ypt7 guanine nucleotide exchange factor complex Mon1-Ccz1. Nat. Commun. 8, 14034 (2017).
Chen, D. C., Yang, B. C. & Kuo, T. T. One-step transformation of yeast in stationary phase. Curr. Genet. 21, 83–84 (1992).
Green M. R., Sambrook J., Sambrook J. Molecular Cloning: a Laboratory Manual, 4th edn. Cold Spring Harbor Laboratory Press (2012).
Cole, J. L., Lary, J. W., Moody, T. P. & Laue, T. M. Analytical ultracentrifugation: Sedimentation velocity and sedimentation equilibrium. Methods Cell. Biol. 84, 143–179 (2008).
Schuck, P. Size-distribution analysis of macromolecules by sedimentation velocity ultracentrifugation and Lamm equation modeling. Biophys. J. 78, 1606–1619 (2000).
Kabsch, W. Integration, scaling, space-group assignment and post-refinement. Acta Crystallogr. D. Biol. Crystallogr. 66, 133–144 (2010).
McCoy, A. J. et al. Phaser crystallographic software. J. Appl. Crystallogr. 40, 658–674 (2007).
Emsley, P. & Cowtan, K. Coot: model-building tools for molecular graphics. Acta Crystallogr. D. Biol. Crystallogr. 60, 2126–2132 (2004).
Adams, P. D. et al. PHENIX: a comprehensive Python-based system for macromolecular structure solution. Acta Crystallogr. D Biol. Crystallogr. 66, 213–221 (2010).
Afonine, P. V. et al. FEM: feature-enhanced map. Acta Crystallogr. D. Biol. Crystallogr. 71, 646–666 (2015).
Murshudov, G. N. et al. REFMAC5 for the refinement of macromolecular crystal structures. Acta Crystallogr. D. Biol. Crystallogr. 67, 355–367 (2011).
Langer, G., Cohen, S. X., Lamzin, V. S. & Perrakis, A. Automated macromolecular model building for X-ray crystallography using ARP/wARP version 7. Nat. Protoc. 3, 1171–1179 (2008).
Chen, V. B. et al. MolProbity: all-atom structure validation for macromolecular crystallography. Acta Crystallogr. D. Biol. Crystallogr. 66, 12–21 (2010).
Zheng, H. P. et al. Validation of metal-binding sites in macromolecular structures with the CheckMyMetal web server. Nat. Protoc. 9, 156–170 (2014).
The PyMOL molecular graphics system, v. 1.8. (Schrödinger LCC, New York, 2015).
Humphrey, W., Dalke, A. & Schulten, K. VMD: Visual molecular dynamics. J. Mol. Graph. Model. 14, 33–38 (1996).
Best, R. B. et al. Optimization of the additive CHARMM All-atom protein force field targeting improved sampling of the backbone phi, psi and side-chain chi(1) and chi(2) dihedral angles. J. Chem. Theory Comput. 8, 3257–3273 (2012).
Essmann, U. et al. A smooth particle mesh ewald method. J. Chem. Phys. 103, 8577–8593 (1995).
Phillips, J. C. et al. Scalable molecular dynamics with NAMD. J. Comput. Chem. 26, 1781–1802 (2005).
Bakan, A. et al. Evol and ProDy for bridging protein sequence evolution and structural dynamics. Bioinformatics 30, 2681–2683 (2014).
Bakan, A., Meireles, L. M. & Bahar, I. ProDy: protein dynamics inferred from theory and experiments. Bioinformatics 27, 1575–1577 (2011).
This work was performed in the framework of SFB 1035 (German Research Foundation DFG, Sonderforschungsbereich 1035, projects A02, B05, and B12). Computational resources for this project were provided by the Gauss Centre for Supercomputing/Leibniz Supercomputing Centre (grant: pr84pa). We thank the staff of the beamline X06SA at the Paul Scherrer Institute, Swiss Light Source for their assistance in X-ray data collection. Dan Humphreys is acknowledged for critically commenting on the manuscript. Christine John and Maximilian Biebl are acknowledged for analytical ultracentrifugation experiments.
The authors declare no competing financial interests.
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
About this article
Cite this article
Wachtel, R., Bräuning, B., Mader, S.L. et al. The protease GtgE from Salmonella exclusively targets inactive Rab GTPases. Nat Commun 9, 44 (2018). https://doi.org/10.1038/s41467-017-02110-1