Characterization of Enzymes from Legionella pneumophila Involved in Reversible Adenylylation of Rab1 Protein*

Background: Covalent modification of small GTPases by pathogens is an emerging field in current research. Results: Quantitative analysis of effects, kinetics, and substrate specificities of adenylylation by DrrA and deadenylylation by SidD was performed. Conclusion: Adenylylation and deadenylylation are means to tightly regulate Rab1 function by Legionella proteins. Significance: This study increases our understanding of Legionella pneumophila subverting Rab protein function during infection. After the pathogenic bacterium Legionella pneumophila is phagocytosed, it injects more than 250 different proteins into the cytoplasm of host cells to evade lysosomal digestion and to replicate inside the host cell. Among these secreted proteins is the protein DrrA/SidM, which has been shown to modify Rab1b, a main regulator of vesicular trafficking in eukaryotic cells, by transfer of adenosine monophosphate (AMP) to Tyr77. In addition, Legionella provides the protein SidD that hydrolytically reverses the covalent modification, suggesting a tight spatial and temporal control of Rab1 function by Legionella during infection. Small angle x-ray scattering experiments of DrrA allowed us to validate a tentative complex model built by combining available crystallographic data. We have established the effects of adenylylation on Rab1 interactions and properties in a quantitative way. In addition, we have characterized the kinetics of DrrA-catalyzed adenylylation as well as SidD-catalyzed deadenylylation toward Rab1 and have determined the nucleotide specificities of both enzymes. This study enhances our knowledge of proteins subverting Rab1 function at the Legionella-containing vacuole.

After the pathogenic bacterium Legionella pneumophila is phagocytosed, it injects more than 250 different proteins into the cytoplasm of host cells to evade lysosomal digestion and to replicate inside the host cell. Among these secreted proteins is the protein DrrA/SidM, which has been shown to modify Rab1b, a main regulator of vesicular trafficking in eukaryotic cells, by transfer of adenosine monophosphate (AMP) to Tyr 77 . In addition, Legionella provides the protein SidD that hydrolytically reverses the covalent modification, suggesting a tight spatial and temporal control of Rab1 function by Legionella during infection. Small angle x-ray scattering experiments of DrrA allowed us to validate a tentative complex model built by combining available crystallographic data. We have established the effects of adenylylation on Rab1 interactions and properties in a quantitative way. In addition, we have characterized the kinetics of DrrA-catalyzed adenylylation as well as SidD-catalyzed deadenylylation toward Rab1 and have determined the nucleotide specificities of both enzymes. This study enhances our knowledge of proteins subverting Rab1 function at the Legionella-containing vacuole.
Eukaryotic cells contain multiple different membrane enclosed compartments or organelles with varying chemical environments. For propagation and survival of the cells, a constant exchange of material between these organelles is necessary, and this is achieved by fission and fusion of vesicles from donor and with acceptor membranes, respectively. Rab proteins, which belong to the superfamily of small GTPases, are the main regulators of this vesicular trafficking in eukaryotic cells (1)(2)(3). During vesicular transport processes, Rab proteins undergo nucleotide exchange between GDP-(inactive state) and GTP-bound (active state) conformations. This conversion is catalyzed by GTPase-activating proteins (GAPs) 7 that catalyze the hydrolysis of bound GTP and guanine nucleotide exchange factors (GEFs) that catalyze GDP-GTP exchange leading to activation of Rab proteins. These regulatory proteins achieve a tight temporal and spatial control of Rab activation and inactivation in eukaryotes (4 -7).
Macrophages use vesicular trafficking pathways to ensure degradation of phagocytosed bacteria. After the bacteria are taken up by the cells via phagocytosis, early and late endosomes fuse with the bacteria-containing vacuoles. During this lysosomal degradation pathway, hydrolytic enzymes are enriched and the bacterial environment is acidified, ensuring degradation of the enclosed bacteria (8,9). Some bacteria have developed intriguing strategies to circumvent degradation inside macrophages and even multiply inside these cells. The bacterium Legionella pneumophila is a prominent example, and a large amount of research has been performed on the mechanisms and proteins involved in evasion of lysosomal degradation (10). After Legionella is taken up by the cell into a Legionella-containing vacuole (LCV), the bacterium injects more than 250 proteins into the host cell using a type IV secretion system termed DOT/ICM (defect in organelle trafficking/intracellular multiplication) (11,12). Several of these proteins have been observed to modify Rab1 function, i.e. of a Rab protein involved in regulation of vesicular trafficking between the endoplasmic reticulum and the Golgi apparatus, thereby recruiting endoplasmic reticulum-derived vesicles to the LCV and establishing it as a replication vacuole (10). The first protein from Legionella reported to directly modify Rab1 function was DrrA (defect in Rab recruitment A), which was originally described as a protein containing GDP dissociation inhibitor (GDI) displacement factor and GEF activity (13,14). Later it was demonstrated that it contains only GEF activity in a central domain containing residues 340 -533, but that no active displacement (GDI displacement factor activity) occurs (15). The C-terminal domain of the protein was reported to bind phosphatidylinositol-4-phosphate with high affinity (16,17), and the N-terminal domain was shown to be an adenylyltransferase catalyzing the covalent attachment of an AMP moiety to Tyr 77 in the switch II region of Rab1 (18). It was shown that the modifications has only small effects on DrrA-mediated nucleotide exchange, whereas catalysis by the GAPs LepB and TBC1D20 is strongly inhibited. Additionally, the human effector protein Mical-3 is not capable of binding to Rab1 in the modified state, whereas in contrast, the Legionella Rab supereffector LidA binds both Rab1 and adenylylated Rab1 (AMP-Rab1) with high affinity (18 -20). The necessity of a protein catalyzing the demodification of Rab1 has been proposed after identification of the adenylyltransferase domain of DrrA (18). Confirming this hypothesis, recently a protein named SidD from Legionella was demonstrated to catalyze the hydrolysis of AMP-Rab1 (21,22).
Here we report the characterization of the biochemical effects of DrrA-catalyzed modification of Rab1 on the interaction with different proteins in a quantitative manner. We determined the kinetic parameters of adenylylation by DrrA and deadenylylation by SidD and investigated the nucleotide preferences of these enzymes. Additionally, we present a structural model of full-length DrrA containing three distinct domains and corroborate the model by small angle x-ray scattering experiments.
SAXS Measurements and Modeling-Small angle x-ray scattering (SAXS) data were collected at the European Molecular Biology Laboratory (EMBL) X33 beamline on the storage ring DORIS III (Deutsches Elektronen-Synchrotron (DESY)) (25). Protein buffer consisted of 20 mM Hepes, pH 8.0, 50 mM NaCl, 1 mM MgCl 2 , 2 mM dithioerythritol. Samples were measured at 20°C at a minimum of three solute concentrations, ranging from 2 to 7.5 mg/ml. The data were recorded using a 1-M PILATUS detector (DECTRIS) at a sample-detector distance of 2.7 m and a wavelength of 1.5 Å, covering the range of momentum transfer 0.012 Å Ϫ1 Ͻ s Ͻ 0.6 Å Ϫ1 (here, s ϭ 4/sin/, where 2 is the scattering angle). No measurable radiation damage was detected on comparison of four successive time frames with 30-s exposures. The data were processed with ATSAS package (26,27) using standard procedures, averaged after normalization to the intensity of the transmitted beam, corrected for buffer contribution, and extrapolated to infinite dilution using the program PRIMUS (28).
The buffer-subtracted data were extrapolated to infinite dilution and further used for analysis and modeling. The radius of gyration (R g ) was determined from the Guinier approximation (29). Maximum complex dimensions D max and the interatomic distance distribution functions P(r) were calculated using GNOM (30).
Ten ab initio models generated using DAMMIN (31) and GASBOR (32) were averaged with DAMAVER (33), which provides a value of normalized spatial discrepancy. Normalized spatial discrepancy values close to 1 indicate that the models are similar. The reference GASBOR model was aligned with the theoretical x-ray model using Supcomb13 (34). Evaluation of the theoretical scattering curves was done using x-ray-based model of full-length DrrA and fitting to the experimental scattering data was performed using CRYSOL (61,62).
Enzyme Kinetics of DrrA and SidD-Both adenylylation by DrrA and deadenylylation by His 6 -SidD 36 -507 were measured using the change in tryptophan fluorescence as reported previously (35) in a FluoroMax-3 spectrophotometer (HORIBA Jobin Yvon) with excitation at 297 nm and emission at 340 nm at 25°C (buffer: 20 mM Hepes, pH 7.5, 100 mM NaCl, and 5 mM MgCl 2 ). For all adenylylation experiments, the quadruple mutant DrrA fl (N451A/R453A/D480A/S483A) was used to exclude effects of the GEF activity of DrrA (15,24), and all measurements of DrrA-catalyzed adenylylation were performed in the presence of 2.5 units of yeast inorganic pyrophosphatase (New England Biolabs) to degrade pyrophosphate produced in the reaction.
Data fitting of hyperbolic curves plotting initial velocities (v) divided by the enzyme concentration ([E 0 ]) against substrate concentrations ([S]) was performed using the following equation.
Data fitting of sigmoidal curves plotting initial velocities (v) divided by the enzyme concentration ([E 0 ]) against substrate concentrations ([S]) was performed using the following equation For determination of Michaelis-Menten parameters from single progress curves, we used KinTek Explorer (36,37) and the following models for adenylylation by DrrA and deadenylylation by SidD (nucleotidylylated Rab1 is abbreviated NMP-Rab1).

GAP and GEF Assays and Determination of Intrinsic Nucleotide Exchange and GTP Hydrolysis Rates-GAP assays for
LepB and TBC1D20 using mantGTP-loaded Rab1 fl and AMP-Rab1 fl were done in a stopped flow apparatus (Applied Photophysics) with excitation at 365 nm and a 395-nm cut-off filter at 25°C in 20 mM Hepes, pH 7.5, 50 mM NaCl, and 5 mM MgCl 2 .
Nucleotide exchange was measured with mantdGDP-loaded Rab1b. Fast measurements in the presence of DrrA 340 -533 were performed using a stopped flow apparatus (Applied Photophysics) with excitation at 366 nm. Emission was detected using a 420-nm cut-off filter. Long term measurements without enzyme were performed using a FluoroMax-3 (HORIBA Jobin Yvon) with excitation at 360 nm and emission measurements at 440 nm with data collection at 20-s intervals. All measurements were performed in 25 mM Hepes, pH 8, 50 mM NaCl, 5 mM MgCl 2 , and 5 mM dithioerythritol at 25°C.
For data analysis of hyperbolic curves plotting observed rate constants from single exponential fits of the progress curves (k obs ) against enzyme concentrations ([E]), the following equation was used.
Isothermal Titration Calorimetry (ITC) Measurements-Interactions of Mical-3  and Rab1 fl ⅐GppNHp or AMP-Rab1 fl ⅐GppNHp were measured using an ITC 200 microcalorimeter (MicroCal). Measurements were carried out in 20 mM Hepes, pH 7.5, 50 mM NaCl, 2 mM MgCl 2 at 25°C. 500 M Mical-3  in the syringe was injected into the cell containing either 50 M Rab1 fl ⅐GppNHp or 50 M AMP-Rab1 fl ⅐GppNHp. Analysis of the data was performed using the Origin 7.0 Software provided by the manufacturer (MicroCal, LLC ITC).

RESULTS
Biochemical Effects of Rab Adenylylation-We have previously demonstrated that adenylylation of Rab1 in the switch II region has dramatic effects on interactions with GAPs and the human effector protein Mical, whereas the GEF activity of DrrA was only slightly inhibited (18). We have now characterized these effects further by determining equilibrium dissociation constants and enzyme parameters.
Independent K m and k cat values for GAP catalysis by LepB and TBC1D20 were determined for unmodified Rab1, whereas only catalytic efficiencies (k cat /K m ) could be measured for AMP-Rab1 (Fig. 1, A and B, Table 1). The catalytic efficiencies of LepB and TBC1D20 toward AMP-Rab1 are decreased by factors of 1000 and 20, respectively.
Interestingly, although the overall catalytic efficiency (given by k cat /K m ) of the DrrA GEF domain is only reduced by a factor of ϳ3 by the covalent modification of Rab1, the K m value is almost an order of magnitude higher for AMP-Rab1 than for Rab1 (Fig. 1C, Table 1), but this is partly compensated for by a higher k cat value with modified AMP-Rab1.
In addition to the catalytic efficiencies of GAPs and GEFs, we also determined the affinity of the human effector protein Mical-3 (38) for Rab1 using ITC. For unmodified Rab1, a K D value of 2.6 M was observed, whereas no binding could be detected in the case of AMP-Rab1, indicating at least 2 orders of magnitude lower affinity (Fig. 1D, Table 1).
In contrast to the interactions of AMP-Rab1 with specific interaction partners, the intrinsic rates of GDP-GTP exchange and GTP hydrolysis were not significantly affected by the covalent modification with AMP ( Fig. 1, E and F). Intrinsic GTP hydrolysis rate constants of 4.3 ϫ 10 Ϫ5 s Ϫ1 and 5.0 ϫ 10 Ϫ5 s Ϫ1 for Rab1 and AMP-Rab1, respectively, and an intrinsic mantdGDP dissociation rate constant of 2.9 ϫ 10 Ϫ5 s Ϫ1 for both Rab1 and AMP-Rab1 were determined.
A Structural Model of DrrA-Since the protein DrrA was identified as a factor that can subvert Rab1 function in Legionella-infected cells, multiple crystal structures of the different domains of DrrA, but not the full-length protein, have been reported (15,16,18,39,40). The N-terminal adenylyltransferase domain (residues 1-339) is structurally not well characterized because only fragments containing residues 9 -218 (Protein Data Bank (PDB) ID 3NKU) and 193-550 (PDB ID 3L0I) were successfully crystallized. Because further crystalliza-tion attempts of bigger fragments of the N-terminal domain failed in our hands, we used multiple known crystal structures of DrrA to create a model of full-length DrrA in complex with Rab1 ( Fig. 2A). However, due to the small overlapping region of fragments 9 -218 and 193-550, the relative orientation of the N-terminal fragments could only be modeled based on the high structural homology with E. coli glutamine synthetase adenylyltransferase (PDB ID 1V4A) (18,41).  (Fig. 2B, indicated by a red arrow) is quite far from the GEF binding site of Rab1, and on the opposite side of DrrA (Fig. 2B), in harmony with the idea of completely different binding modes for catalysis of nucleotide exchange and adenylylation of Rab1 by DrrA.
The hypothetical model was further validated using SAXS. The analysis of the Guinier plot indicated the absence of strong interparticle interactions, such as repulsion or attraction, and a radius of gyration (R g ) of 3.9 Ϯ 0.2 nm (supplemental Fig. S1). The maximum particle dimensions (D max ) and the interatomic distance distribution function P(r) were estimated using GNOM (30). The P(r) function (Fig. 2C) showed a somewhat increased D max value (15.5 Ϯ 0.5 nm) as compared with the theoretical model (D max ϭ 14.6 nm), which can be attributed to the missing termini in the theoretical model (missing residues  (15)). Data fitting for A-C (see "Experimental Procedures") led to the k cat , K m , and k cat /K m values shown in Table 1. D, ITC measurements for interaction of Mical-3  with Rab1 fl ⅐GppNHp (left, n ϭ 0.880 Ϯ 0.011, ⌬H ϭ Ϫ7772 Ϯ 162 cal/mol, ⌬S ϭ Ϫ0.536 Ϯ cal/mol/K), and AMP-Rab1 fl ⅐GppNHp (right). E and F, in contrast to interaction with different proteins, intrinsic GTP hydrolysis (k obs ϭ 4.3 ϫ 10 Ϫ5 s Ϫ1 for Rab1; k obs ϭ 5.0 ϫ 10 Ϫ5 s Ϫ1 for AMP-Rab1) (E) and intrinsic nucleotide exchange rates (k obs ϭ 2.9 ϫ 10 Ϫ5 s Ϫ1 for Rab1 and AMP-Rab1) (F) were not affected by the covalent modification.

TABLE 1 Enzyme parameters and equilibrium dissociation constants
The table summarizes the determined enzyme parameters (GAPs LepB and TBC1D20, GEF DrrA) and equilibrium binding constants of the human effector protein Mical for Rab1 and AMP-Rab1, respectively. -indicates not detectable or applicable. 1-15 and 640 -647) or a more extended conformation of the complex in solution. The theoretical pattern calculated from the x-ray-based model can be well fitted to the experimental data with a discrepancy 2 value of ϳ1 (supplemental Fig. S1A). Ab initio modeling using DAMMIN identified an excluded volume of 175 Ϯ 15 nm 3 , which is in good agreement with the expected molecular mass of the complex (ϳ95 kDa). Normalized spatial discrepancy values were below 0.7, indicating that independent reconstructions converged to a similar shape. Higher resolution ab initio models, consistent with the expected molecular mass, were built using GASBOR. The reference GASBOR model was aligned to an x-ray-based model, showing a good agreement with the latter (Fig. 2B), hence suggesting that the x-ray-based model is a close representation of the complex in solution. However, one has to keep in mind that a membranous environment and additional, yet to be identified protein interactions might influence the conformation of DrrA.

Protein
Kinetics of DrrA-catalyzed Adenylylation-First measurements of adenylylation by DrrA using high concentrations of Rab1⅐GppNHp as substrate showed a fast initial rate of low amplitude followed by a linear phase in the progress curve (supplemental Fig. S2). This kinetic behavior was not easily explainable by simple competitive inhibition, but rather by allosteric inhibition by a reaction product affecting the turnover number (k cat ) of DrrA. Because the generated pyrophosphate was a possible cause of this product inhibition, we repeated the measurement in the presence of pyrophosphatase, which led to a more rapid overall time course that could be well fitted based on a simple Michaelis-Menten model without product inhibition. Therefore, all experiments for determination of Michaelis-Menten parameters were performed in the presence of pyrophosphatase. Interestingly, systematic analysis of the dependence of the initial rates on substrate concentrations showed sigmoidal kinetic behavior for Rab1⅐GDP, but not Rab1⅐ GppNHp, using ATP as the cosubstrate. The sigmoidal fit resulted in a Hill parameter of 1.7, indicating possible dimer formation of Rab1⅐GDP at high concentrations as was shown for Rab9 and Rab27 (42,43). Interestingly, the switch II region of the dimeric GDP-bound Rab9 was reported to adopt an active-like conformation, which might explain the increased reaction rates of adenylylation toward Rab1⅐GDP at higher concentrations (43). However, because the inactive GDP-bound state of Rab1 is not a physiologically relevant substrate, the sigmoidal kinetic behavior was not further investigated at this point.
Confirming the previously reported preference of DrrA-catalyzed adenylylation toward Rab1⅐GppNHp as compared with  Table 2). Because of the relatively low K m value for Rab1⅐GppNHp, calculation of independent k cat and K m values was also possible from a single progress curve, and the values obtained were in good agreement with the previously mentioned parameters (supplemental Fig.  S3A). In summary, these findings indicate a 395-fold higher catalytic efficiency of DrrA toward the active state Rab1⅐ GppNHp as compared with the inactive Rab1⅐GDP. Furthermore, a K m value of 73.9 M was measured for ATP ( Fig. 3C and Table 2). Because determination of the catalytic parameters for ATP was performed at nonsaturating concentrations of Rab1⅐GppNHp, the corresponding k cat and k cat /K m values are only apparent values and are shown in parentheses ( Table 2).
Kinetics of SidD-catalyzed Deadenylylation-Experiments using cell lysates from both E. coli and L. pneumophila showed a specific activity for removal of the AMP-modification in Legionella, but not in E. coli lysates (supplemental Fig. S4), confirming the presence of the previously identified deadenylylase SidD (21,22) and indicating the necessity of this specific enzyme for removal of the covalent modification. To obtain insights into the competition of adenylylation by DrrA and deadenylylation by SidD, we quantified the catalytic activity of recombinantly produced and purified SidD. Surprisingly, deadenylylation catalyzed by SidD was not significantly affected by the activation state of Rab1 (Fig. 4). K m values of 2.6 and 3.9 M and k cat values of 13.1 and 17.8 s Ϫ1 for AMP-Rab1⅐GDP and AMP-Rab1⅐GppNHp, respectively, were determined ( Table 3). Because of the low K m values, the calculation of independent k cat and K m values was possible from single progress curves (supplemental Fig. S3B), indicating good agreement with the parameters obtained from Michaelis-Menten plots.
Nucleotide Specificity of DrrA and SidD-In addition to ATP, DrrA was also reported to be capable of using GTP as substrate for guanylylation of Rab1 (18). Hence, we systematically analyzed the substrate properties of various nucleotides. Besides ATP and GTP, DrrA could also utilize ADP and GDP, albeit at lower rates (Fig. 5A). Furthermore, pyrimidine nucleotides (CTP and UTP) were also tested. Interestingly, CTP showed similar substrate properties to ATP, whereas UTP showed only very low reaction rates (Fig. 5B), indicating an important role of the amino group at the C6 position of the purine ring and the C4 position of the pyrimidine ring, respectively, in recognition by DrrA. Determination of the initial velocities of the nucleotidylylation reaction for each nucleotide indicated 60% activity for CTP, ϳ14% for ADP and GTP, and less than 10% for the remaining nucleotides tested as compared with ATP (Fig. 5C).
Because DrrA was capable of modifying Rab1 with various nucleotides, we consequently tested the reversibility using SidD in a fluorescence-based assay (Fig. 5E) and confirmed the results using electrospray mass ionization-MS at the end of each reaction (data not shown). For all nucleotide modifications catalyzed by DrrA, SidD catalyzed the demodification, albeit with slightly lower catalytic efficiencies for GMP, CMP, and UMP than for AMP ( Fig. 5E and Table 4). The experiments show that ATP and AMP-Rab1 are the preferred substrates for DrrA and SidD, respectively, but other nucleotides might be used during infection as well.

DISCUSSION
During the past few years, posttranslational modifications and especially adenylylation of host cell proteins by pathogenic bacteria have become a topic of great interest (18, 44 -50). Very recently, it has also been demonstrated for DrrA-catalyzed adenylylation that the modification can be reversed by the
Here, we present a model giving deeper insight into the DrrA-and SidD-catalyzed events taking place at the LCV including knowledge on the kinetics of adenylylation by DrrA and deadenylylation by SidD determined in this study and data on the time-dependent detection of different Legionella proteins at the LCV reported in literature. We speculate that the infection can most probably be divided into two phases regarding subversion of Rab1 function by Legionella (Fig. 6).
The first phase is referred to as the entrapment phase in Fig.  6. This phase is characterized by adenylylation of Rab1, and both DrrA and Rab1 are localized at the LCV (52). The adenylylation of Rab1 might be used as a tool to interfere with the control of Rab1 by the host cell, such as for example inactivation by GAPs. Furthermore, the covalent modification has been shown to inhibit binding of GDI, thereby not allowing extraction from the LCV membrane (24). During this entrapment phase, the supereffector LidA might act as a tethering factor for endoplasmic reticulum-derived vesicles because it is still capable of binding AMP-Rab1 with high affinity and is present at the LCV within 20 min after infection (20,53). In contrast to LidA from Legionella, binding of the human effector protein Mical is not possible in the modified state. Further studies showing whether the covalent modification inhibits human effector proteins in general from binding to modified Rab1 would be of great interest. The absence of Rab1 effector proteins p115 and GM130 at the LCV during infection (54) indicates decreased binding affinities of these proteins to AMP-Rab1 as well.
The preference of DrrA-catalyzed adenylylation toward Rab1⅐GppNHp and the catalytic efficiency of the DrrA GEF domain of 2.2 ϫ 10 5 M Ϫ1 s Ϫ1 (15) suggest a consecutive order of DrrA function. After binding to the LCV via the P4M domain, DrrA recruits Rab1 by catalyzing nucleotide exchange, and only afterward, Rab1 can be covalently modified with AMP. This is further supported by our model of DrrA fl , which we corroborated by small angle x-ray scattering as the localization of the active sites does not allow a simultaneous catalysis or even substrate channeling, but rather different events of Rab binding and release for nucleotide exchange and adenylylation by DrrA.
The second phase (referred to as the release phase in Fig. 6) starts approximately 6 h after internalization, when Legionella begins to replicate (53,55,56). Approximately at this time, DrrA loses its LCV localization (52), whereas both LepB and SidD can be detected at the LCV (21,52). The delayed detection of SidD and LepB in these studies might indicate a generally lower expression level or a delayed transcription and secretion of these proteins as compared with DrrA. While binding of DrrA to the LCV membrane depends on the presence of phosphatidylinositol-4-phosphate, LepB most likely binds via putative transmembrane regions and might be enriched at the LCV during the course of infection even at low expression levels. As discussed later, a putative membrane binding domain has not been identified for SidD yet, and the distribution of SidD inside the host cell during infection is unknown.
Because LepB cannot catalyze GTP hydrolysis on AMP-Rab1, the order of events must be deadenylylation of AMP-Rab1 by SidD followed by deactivation by LepB. Only at this point is GDI able to extract Rab1 from the LCV membrane so that Rab1 is removed from the LCV (24,52).
Because adenylylation by ectopically expressed DrrA has been shown to have cytotoxic effects in cell culture experiments (14,18), demodification and release of Rab1 from the LCV might be necessary to ensure survival of the host cell as long as Legionella replicates. We show that SidD has a high catalytic activity for both AMP-Rab1⅐GDP and AMP-Rab1⅐GppNHp. The high activity might indicate that SidD has evolved to ensure rapid and complete demodification of Rab1 at the LCV as well as other places inside the cell. Even in the presence of small remaining concentrations of DrrA, the high catalytic efficiency would ensure complete demodification. In this respect, it is of great interest to obtain more insight into the local concentrations of the enzymes DrrA and SidD and their substrates at the LCV during infection. The restriction to lateral diffusion for membrane-bound proteins, which effectively increases the  local concentrations, might significantly influence the catalysis in an in vivo environment.
As an explanation of the ready acceptance of the GDP-bound form of AMP-Rab1 by SidD, AMP-Rab1⅐GDP, might adopt an active-like conformation of the switch II region as the crystal structure of AMP-Rab1b⅐GppNHp (PDB ID 3NKV) (18) showed a stacking interaction of the adenosine and Phe 45 , and this interaction might also occur in AMP-Rab1⅐GDP. This   (Table 4). F, the structural formulae of nucleobases are shown in the order of substrate specificity of SidD for Rab1 modified with nucleoside monophosphates.
Characterization of adenylylating Enzymes from L. pneumophila OCTOBER 12, 2012 • VOLUME 287 • NUMBER 42 could therefore explain the low substrate specificity of SidD toward GDP or GTP/GppNHp-bound AMP-Rab1. Alternatively, SidD might solely recognize the amino acid sequence rather than any secondary structure of the switch II region of Rab1.
Although DrrA can use ATP, GTP, UTP, and CTP for nucleotidylylation of Rab proteins, the determined catalytic parameters suggest a preference for ATP. This specificity is likely to be even more pronounced in vivo because the intracellular nucle-  57)). Therefore, these concentrations in conjunction with our biochemical nucleotidylylation parameters suggest ϳ91% modification of Rab1 with AMP, less than 6% modification of Rab1 with CMP, and less than 3% modification of Rab1 with GMP, respectively, during infection. Modifications with UMP are most probably insignificant. This might therefore also explain the requirement that SidD is able to remove different covalently attached nucleotides. Supporting the preference of ATP for modification of Rab1, mass spectrometric analyses of cells infected with L. pneumophila detected AMP-Rab1, but no other nucleotide-modified Rab1 species (58).
Because SidD contains 507 amino acids, it is probably a multidomain protein. Identification and modeling of a protein phosphatase fold in SidD indicates N-terminal localization of the deadenylylation domain (59), raising the question of potential functions of the C terminus of SidD. Further characterization of the domain architecture as well as characterization of putative other domains will be of great interest. In this regard, the presence of a putative lipid binding domain for localization of SidD, as was shown for multiple other Legionella secreted proteins (12), and the intracellular localization of SidD during infection will give further insight into its purpose during infection at the LCV and possibly other places inside the cell.
After characterization of the antagonists AnkX and Lem3/ lpg0696 for Rab1 modification (24,35), this study gives further insight into the role of DrrA and SidD during infection. It is of great interest for future research to determine whether pathogens using Fic domain-containing proteins for adenylylation of host cell GTPases (46,60) have evolved similar antagonistic proteins to reverse the modification in a time-controlled manner during infection.
In this study, we have reported a model of the order of events taking place at the LCV based on the detailed characterization of DrrA and SidD and including known factors of Rab1 interaction. Looking at the amount of proteins secreted by Legionella with yet unknown function and the recent progress in identification of further Rab-interacting and -modifying proteins, identification of more factors for further fine tuning of Rab protein function and subversion of vesicular transport will be an interesting issue for future research.