Rab1-AMPylation by Legionella DrrA is allosterically activated by Rab1

Legionella pneumophila infects eukaryotic cells by forming a replicative organelle – the Legionella containing vacuole. During this process, the bacterial protein DrrA/SidM is secreted and manipulates the activity and post-translational modification (PTM) states of the vesicular trafficking regulator Rab1. As a result, Rab1 is modified with an adenosine monophosphate (AMP), and this process is referred to as AMPylation. Here, we use a chemical approach to stabilise low-affinity Rab:DrrA complexes in a site-specific manner to gain insight into the molecular basis of the interaction between the Rab protein and the AMPylation domain of DrrA. The crystal structure of the Rab:DrrA complex reveals a previously unknown non-conventional Rab-binding site (NC-RBS). Biochemical characterisation demonstrates allosteric stimulation of the AMPylation activity of DrrA via Rab binding to the NC-RBS. We speculate that allosteric control of DrrA could in principle prevent random and potentially cytotoxic AMPylation in the host, thereby perhaps ensuring efficient infection by Legionella. The Legionella effector DrrA AMPylates the host protein Rab1 during infection, but the mechanism is still under debate. Here, the authors provide structural insights into the low-affinity DrrA:Rab1 interaction, showing that Rab1 allosterically activates DrrA through a non-conventional binding mechanism.

T he Gram-negative bacterium Legionella pneumophila is the causative agent of Legionnaires' disease. After uptake by human alveolar macrophages via phagocytosis, the pathogen establishes a replicative organelle referred to as the Legionella-containing vacuole (LCV) 1 . The formation and maintenance of the LCV are mediated by~330 Legionella effector proteins that are released by the bacterial Type IVb secretion system (T4bSS) from the bacterium into the host. These effectors interfere with different host processes, including manipulation of vesicular trafficking (summarised in ref. 2 ). In particular, one intensely studied protein, Legionella effector DrrA (Defect in Rab1 recruitment A, also referred to as SidM, substrate of Icm/ Dot), is secreted early during infection, and it manipulates the vesicular trafficking regulator Rab1 (Ras-related protein Rab-1).
Rab1 is a member of the Rab family of small GTPases that are involved in spatially and temporally regulating intracellular vesicular trafficking between organelles. Rab proteins act as molecular switches that exist in inactive guanosine diphosphate (GDP) and active guanosine triphosphate (GTP) bound states 3 . At the structural level, the activity state is communicated to interaction partners via two important regulatory loop regions referred to as switch I and switch II. These regions are conformationally flexible in the inactive state but they become structurally ordered in the active form 4 . Only in the active form, Rab proteins mediate signalling via the recruitment of GTP statespecific effector molecules. Guanine nucleotide exchange factors (GEFs) promote GTP loading of Rab, whereas GTPase activating proteins (GAPs) stimulate the intrinsic GTP hydrolysis activity and return Rab back to the inactive state 3 . Furthermore, Rab proteins are post-translationally modified with geranylgeranyl lipids at the structurally flexible C-terminus, enabling reversible membrane attachment 5 . The recycling of geranylgeranylated Rabs from the membrane occurs in the inactive state and is controlled by GDP dissociation inhibitors (GDIs). The biology of Rab proteins has been the topic of a recent review 3 .
Due to their pivotal role in regulating vesicular trafficking, Rab proteins are frequently targeted and manipulated by bacterial pathogens. Rab1 controls vesicular transport from the endoplasmic reticulum (ER) to the Golgi apparatus 6,7 . However, it is subjected to intense manipulation during Legionella infections and is rerouted from the ER to the LCV. In this process, Rab1 is known to be targeted by six different Legionella proteins, one of which is DrrA 2,8,9 . However, the deletion of DrrA has little impact on successful infection by Legionella, probably due to effector redundancy 9 .
DrrA consists of three functional domains (Fig. 1a). The Cterminal phosphatidylinositol-4-phosphate (PI4P) binding site (referred to as P4M) mediates binding to the PI4P-positive LCV membrane 10,11 . A central Rab1-GEF domain recruits Rab1 via GDP-to-GTP exchange 8,9,12 . The N-terminal domain possesses AMPylation activity (also referred to as adenylylation) 13 . AMPylation is a post-translational modification (PTM) in which an adenosine triphosphate (ATP) is utilised in order to transfer adenosine monophosphate (AMP) to hydroxyl-bearing amino acid side chains. The N-terminal AMP-transferase (ATase) activity of DrrA preferentially mediates AMPylation of active Rab1b at Y77 Rab1b (i.e. Y80 Rab1a in Rab1a), which is located at the C-terminal end of the important Rab1 switch II region 13,14 . AMPylation inhibits binding to interaction partners such as GDI and molecule interacting with CasL 3 (MICAL3). Also, GAP-catalysed GTP hydrolysis by Rab1 is impaired, thereby rendering the protein permanently bound to GTP when AMPylated 13,15 . Notably, it was reported that GEF-deficient DrrA mutants are competent in the recruitment of Rab1 from Rab1:GDI complex to the LCV 16 .
The structure of DrrA has been partially characterised [11][12][13]17 . However, the structure of the full ATase domain and the mechanism of Rab1b AMPylation remain elusive. The ATase domain belongs to the DNA polymerase β-like enzyme family and shares structural similarity with the enzyme glutamine synthetase adenylyl transferase (GS-ATase) from Escherichia coli 18 . In this class of enzymes, three acidic amino acid residues (D110 DrrA , D112 DrrA and D150 DrrA ) are involved in coordinating an essential Mg 2+ relevant for ATP binding and AMP transfer. The structural basis for AMP transfer and Rab1 binding has not yet been reported, since structural investigation of Rab1-DrrA ATase complexes is hampered by inherently low affinity; the Michaelis constant (K M ) is only~64 µM 14 .
Herein, we employ covalent crosslinking approaches to stabilise and characterise low-affinity DrrA ATase :Rab-complexes. The crystal structure of a DrrA ATase :Rab-complex reveals an unrecognised regulatory Rab binding site. By mutational analysis, mass spectrometry and activity determinations we show that binding of Rab to this particular position on DrrA acts as a safety switch to control the AMPylation activity of DrrA via an allosteric mechanism in vitro. Our findings lead us to speculate that a potentially harmful AMPylation activity of DrrA may be controlled by Rab1:GTP-binding to an allosteric site.

Results
Conceptual design for trapping the DrrA-Rab complex. Since the ATase domain of DrrA binds to Rab1b only with low affinity 14 , we attempted to covalently trap and thereby homogenously enrich and stabilise the complex via two different chemical approaches. We first employed a crosslinking strategy that we recently developed based on site-specific incorporation of unnatural amino acids (UAAs) bearing bromoalkyl moieties (e.g. BrC6K, Fig. 1b) that react specifically with nucleophilic natural amino acids (e.g. Cys, Asp and Glu) in a proximity-enhanced manner upon complex formation [19][20][21] . The main effect of proximity consists in increasing the effective local concentration of the reactants to boost reaction rates and enable reactions that would not yield products in the absence of the concentration effect. To stabilise the DrrA ATase -Rab1b interface, we overexpressed various C-terminally His6-tagged Rab1b constructs (amino acids 3−174; Rab1b 3−174 with a Q67A Rab1b mutation that minimises Rab1b GTP hydrolysis) 22,23 bearing BrC6K at different positions (R69 Rab1b , T72 Rab1b , I73 Rab1b and R79 Rab1b ) in the vicinity of the AMPylation site (Y77 Rab1b ), together with an Nterminally Strep-tagged wild type (wt) DrrA ATase construct (amino acids 16−352; DrrA 16−352 ). Co-expression of Rab1b with BrC6K at position 69 Rab1b and wt DrrA ATase resulted in the formation of a covalently crosslinked complex, as confirmed by sodium dodecyl sulphate polyacrylamide gel electrophoresis (SDS-PAGE) shift analysis via α-His6 and α-Strep western blotting ( Fig. 1c and Supplementary Fig. 1). We purified the crosslinked complex by affinity and size-exclusion chromatography, and performed tandem mass spectrometry (MS/MS), which indicated that the crosslink corresponds to an ester linkage formed between BrC6K at position 69 Rab1b within Rab1b and D82 DrrA within DrrA (Supplementary Fig. 2 and Supplementary  Table 1). This finding was corroborated by mutagenesis experiments, in which the D82A DrrA mutation completely abolished crosslink formation, while mutating the Asp residue with the more nucleophilic Cys led to quantitatively thioether crosslinked Rab1b:DrrA complex ( Fig. 1c and Supplementary Fig. 1). Electrospray ionisation MS (ESI-MS) of the purified complex formed in living E. coli cells between R69BrC6K Rab1b and D82C DrrA verified the identity of the thioether crosslinked complex, which is AMPylated at Y77 Rab1b within Rab1b, confirming the enzyme and substrate activity of both DrrA and Rab1b mutants (Fig. 1d). Intriguingly, amino acid D82 DrrA in DrrA is not positioned in the vicinity of the catalytically important amino acids D110 DrrA and D112 DrrA 13 but rather on a surface patch located opposite to the active site of DrrA ATase , indicating either complete rearrangement of the previously characterised DrrA fold or an alternative mechanism for enzyme activation. Importantly, full-length D82C DrrA including the GEF and P4M domains (amino acids 1 −647; DrrA 1−647 ) also displayed efficient crosslinking upon coexpression of R69BrC6K Rab1b (Supplementary Fig. 1). In order to gain structural insight into Rab1b binding to DrrA ATase , we generated several milligrams of thioether-crosslinked DrrA  : Rab1b 3−174 complex and subjected them to crystallisation trials.
In parallel, we used a complementary in vitro crosslinking approach to stabilise the low-affinity DrrA ATase :Rab interface using our recently developed thiol-reactive nucleotide derivatives (TReNDs) as bridging functionalities 24 (Fig. 1e) ARTICLE to a strategically placed cysteine in a proximity-enabled manner. In this case, chloroacetamide is positioned at an ATP derivative in which the nucleobase is replaced by a triazole ring. Thus, the combined reaction of the chloroacetamide with a strategically introduced cysteine in DrrA and the AMP transfer by DrrA to Rab proteins enables the formation of a covalently trapped complex between the enzyme and its target protein. Inspired by the identification of D82C DrrA as a site for crosslinking with BrC6K-bearing Rab1b (see above) and by a previous prediction of the ATP-binding site in DrrA 25 , we intended to probe both the putative adenine binding pocket and the amino acid residues located at the opposite surface of DrrA for complex formation with TReNDs and Rab1b. Therefore, we generated DrrA 16−352 constructs bearing M169C, M174C, V175C (the putative adenine interaction interface) and L197C, and G198C (opposite to the putative adenine interaction site) substitutions (Fig. 1f), and tested their ability to form a covalent complex with TReNDs and Rab1b 3−174 loaded with the non-hydrolysable GTP analogue GppNHp. TReND-1 was indeed transferred by wt DrrA 16−352 to Rab1b 3−174 , indicating that it is correctly positioned in the catalytic centre ( Supplementary Fig. 3). Using recombinantly purified proteins, TReND-1 successfully formed a covalent ternary complex with Rab1b 3−174 and DrrA 16−352 -L197C, as demonstrated by an apparent increase in molecular weight in denaturing SDS-PAGE (Fig. 1g), while ATP failed to form a covalently linked ternary complex, as expected. The reaction was site-specific since other mutants did not produce a covalent product ( Fig. 1g and Supplementary Fig. 4). Importantly, residue L197 DrrA is also not located in the vicinity of the ATase active site; it is situated on the same surface patch as residue D82 DrrA opposite the catalytically active amino acid residues of DrrA ( Supplementary Fig. 1). Encouraged by the efficiency of our crosslink formation with the DrrA 16−352 -L197C mutant, we attempted to produce larger quantities of the ternary DrrA-TReND-1-Rab1b complex for structural investigations by X-ray crystallography. Several milligrams of pure DrrA 16-352 :TReND-1:Rab1b 3-174 complex were obtained, but the resulting crystals diffracted only poorly as it was the case for the thioether crosslinked DrrA 16-352 :Rab1b 3−174 described above. Therefore, we also generated the TReND-1linked complex with the GTPase domain of the close Rab1bhomologue Rab8a (amino acids 6−176; Rab8a 6-176 ; Fig. 1h). Rab1b and Rab8a share 53% sequence identity, their structures can be superimposed with an RMSD (root-mean-square deviation) of about 0.5 Å, and Rab8a can be AMPylated by DrrA in vitro with very similar kinetics ( Supplementary Fig. 5) 13 . The integrity, homogeneity, and correctness of the covalently linked ternary DrrA:TReND-1:Rab8a complex were validated using highresolution (HR) MS of intact proteins, and the results demonstrated excellent accordance between experimental and theoretical mass (Fig. 1i). Thus, combining thiol-reactive UAAs or nucleotides with cysteine-substituted DrrA allowed us to prepare covalent DrrA-Rab complexes, overcoming the difficulties in accessing the transient DrrA ATase -Rab complex by more traditional approaches.
Structure of the DrrA:Rab8a complex. In order to obtain insight into the molecular basis of DrrA substrate recognition, we determined the crystal structure of the DrrA 16-352 :TReND-1: Rab8a 6-176 complex bound to GppNHp at 2.1 Å resolution (DrrA 16-352 :TReND-1:Rab8a 6-176 will be referred to as DrrA: Rab8a hereafter for clarity) (Fig. 2a, Supplementary Table 2 and Supplementary Fig. 6). The complex binding mode observed in the crystal structure was unexpected since the Rab molecule does not interact with the catalytic centre of DrrA 16-352 constituted by the catalytically important residues D110 DrrA and D112 DrrA 13 . Instead, Rab8a:GppNHp binds to a surface patch opposite the catalytic centre that we refer to as the non-conventional (NC) Rab binding site (RBS), hereafter referred to as NC-RBS (Fig. 2a, d). The complex exhibits the general structural features of the previously determined individual subunits of Rab8a:GppNHp, and parts of DrrA  and DrrA 210-534 (Fig. 2a) 13,17,26 . In accordance with previously solved structures, Rab8a in complex with DrrA has a fold consisting of a central six-stranded β-sheet surrounded by five α-helices 26 . Albeit individual crystal structures of N-and C-terminal parts of DrrA are available 13,17 , the structure of the full ATase domain of DrrA has not been reported to date. The general arrangement of the secondary structure elements of the so far solved fragments agrees very well with the full ATase domain, yet significant positional differences in the catalytic site residues were observed. Based on the structures of other ATases belonging to the DNA polymerase β-like enzyme family, the catalytic motif of DrrA is expected to be G98-S99-L100-X 11 -D110-X-D112 DrrA , with D110 DrrA and D112 DrrA involved in coordinating the essential magnesium ion. In contrast to the previous DrrA 8-218 fragment structure 13 , the position of D110 DrrA and D112 DrrA is well defined and the residues appear to be capable of Mg 2+ coordination (Fig. 2a).
The covalently bound TReND-1 is well defined in the electron density in the complex crystal structure (Fig. 2b). However, due to structural differences of the chemical linker, bridging the TReND-1 ribose and C197 in DrrA, compared to ATP, only few interactions with the central scaffold of the linker are formed, which is reflected by the higher B-factors in respect to the surrounding amino acid residues ( Supplementary Fig. 7). In the DrrA:Rab8a complex structure, switch I (E30 Rab8a , F33 Rab8a and N34 Rab8a ), the interswitch region (D44 Rab8a and K58 Rab8a ), and switch II (Q60 Rab8a and R69 Rab8a ) of Rab8a mainly interact with the N-terminal half of the ATPase domain of DrrA opposite the catalytic motif, centred at amino acids R70 DrrA , Q71 DrrA and K74 DrrA . Evaluation using the Protein Interfaces, Surfaces and Assemblies (PISA) web service 27 showed that the solventaccessible area buried upon complex formation is small (648 Å 2 ), indicating that the complex is of low affinity and requires covalent linkage in order to be observed ( Fig. 2c; hydrophobic interactions can be seen in Supplementary Fig. 8). In the previous structure of the N-terminus DrrA ATase , the catalytic site (D110 DrrA , D112 DrrA and D150 DrrA ) was disordered and incompetent for Mg 2+ coordination in this arrangement. However, in our complex crystal structure, all three acidic amino acids are properly positioned upon Rab binding to the NC-RBS, as observed by superimposition with the ATase-domain of glutamine synthetase adenylyl transferase from E. coli 18 (Fig. 2e). Therefore, the C-terminal half of DrrA ATase appears to be important for overall structural integrity and correct alignment of the catalytic site.
The structure of free Rab8a:GppNHp superimposes very well with Rab8a in the NC-RBS-complex, with an RMSD of 0.38 Å, indicating that binding of Rab8a to DrrA does not result in global structural changes in Rab8a (Fig. 2f) 26 . Similarly, AMPylated Rab1b also superimposes with Rab8a in the NC-RBS-complex, with an RMSD of about 0.6 Å, demonstrating that binding to the NC-RBS of DrrA is also unlikely to induce significant conformational changes in Rab1b (Fig. 2f) 13 .
Consequently, the DrrA:Rab8a crystal structure reveals a previously unrecognised NC-RBS opposite the catalytic centre of the ATase.
Validation of the non-conventional Rab-DrrA interface. In order to rule out the possibility that binding of Rab1b or Rab8a to the NC-RBS is an artefact of covalent linkage caused by TReND-1, ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-20702-2 we validated the observed complex structure using complementary approaches. First, we produced alanine mutants of DrrA amino acids involved in the interface with Rab8a. Since the Rab-proteins can interact with the NC-RBS and the catalytic site of DrrA ATase at the same time (Fig. 2d), a distinction of the effect of Rab-mutations on AMPylation by binding to either site would not be possible. Hence, we only included DrrA alanine mutations for subsequent experiments. To assess whether the alanine substitutions impaired the general stability of DrrA, we subjected them to thermal unfolding and monitored the change in circular dichroism (CD) signal. Compared with the melting point (T m ) of wt DrrA (T m = 60.1°C), the other alanine substitutions, but not D79A DrrA , displayed mild changes with an amplitude from 0.1 to 2.5°C, demonstrating that these mutants are biochemically similar to wt DrrA 16−352 (Fig. 3a). D79 DrrA forms intramolecular polar interactions with K36 DrrA and Y40 DrrA in the complex crystal structure, and D79A DrrA destabilised DrrA, as evidenced by a decrease of 4.2°C in the melting point.
Since the complex was obtained with Rab8a as a Rab1bsurrogate, we tested whether there are notable differences of the mode of DrrA-binding. Structural superimposition of Rab1b with Rab8a from the DrrA:Rab8a complex revealed four amino acids that are located in the interface and differ between the Rabproteins 13 : L26, A29, T32, E35 vs. Rab8a: F26, S29, A32, S35. However, none of these amino acids are involved in interactions in the complex interface and do not create steric clashes that could impair complex formation between DrrA 16-352 and Rab1b ( Fig. 2c and Supplementary Fig. 9). Consistent with the lack of steric clashes and the high sequence homology between Rab1b and Rab8a, the AMPylation rates of these GppNHp-loaded Rabs by DrrA are comparable ( Supplementary Fig. 5). Thus, the mode of binding of DrrA  to Rab1b:GppNHp and Rab8a:GppNHp is identical.
Next, we investigated the AMPylation activity of these DrrA mutants toward GppNHp-bound Rab1b and Rab8a. AMPtransfer results in the decrease of GTPase tryptophan fluorescence and thus can be used for monitoring AMPylation as reported previously 14 . The DrrA mutants D79A DrrA , Y195A DrrA and K74A DrrA exhibited a 5-fold, a 30-fold and a 50-fold decrease in activity toward Rab1, respectively, and R70A DrrA and Q71A DrrA displayed at least 1000-fold activity reduction toward Rab1 ( Fig. 3b and Supplementary Fig. 10). The strong decrease in AMP transfer rates can be explained by an ionic and polar interaction network involving R70 DrrA , Q71 DrrA and K74 DrrA with D44 Rab1b/Rab8a . Although Rab8a was used as a surrogate of Rab1b for determining the complex structure, the kinetics of Rab8a-AMPylation by the DrrA-mutants yields similar results, confirming again an identical mode of DrrA ATase -binding for both Rabproteins (Supplementary Figs. 5 and 10). Therefore, these results suggest that the NC-RBS may play a significant role in regulating DrrA AMPylation activity.
In order to further validate our results, we envisioned testing the complex interface in cellulo using proximity-enabled crosslinking of BrC6K-bearing Rab1b and Cys-mutants of DrrA. Guided by the complex structure, we selected additional combinations for pairwise BrC6K-(within Rab1b) and Cyssubstitutions (within DrrA) that we predicted to result in sitespecific covalent crosslinking when co-expressing the constructs in E. coli. This approach allowed us to screen the interface of the complex within a distance of 13 Å. We confirmed the selective formation of thioether crosslinks between BrC6K incorporated site-specifically at position 69 Rab1b within Rab1b and Cys residues introduced at E75 DrrA , Y78 DrrA and D79 DrrA within DrrA in living E. coli cells using SDS-PAGE, α-His6 and α-Strep westernblotting analyses (Fig. 3c, d and Supplementary Fig. 11). Likewise, crosslinks were observed between D44BrC6K Rab1b and Q71C DrrA , as well as between D31BrC6K Rab1b and E280 DrrA of wt DrrA. Importantly, the pairwise introduction of BrC6K Rab1b and cysteines DrrA at more remote positions (>15 Å) did not result in covalently linked complexes, indicating that the crosslinking  reaction is specific for proximal sites that can be spanned by the flexible bridge formed between BrC6K and nucleophilic residues ( Supplementary Fig. 11).
In summary, our mutational analyses and covalent complex formation studies demonstrate that the interaction of Rab1b with the NC-RBS of DrrA is related to the AMPylation activity of the enzyme, and not an artefact of covalent complex formation using TReND-1 or the UAA BrC6K.
Rab1-binding to the NC-RBS activates DrrA. Even though we confirmed the presence of an NC-RBS in the complex crystal structure, the functional relevance of Rab1b binding to the noncatalytic site remains unclear. We first explored whether Rab1b binding causes structural changes in DrrA. Since the apo structure of DrrA has not yet been determined, superimposition of the previously reported N-and C-terminal parts (DrrA  and DrrA 218-340 , referred as part-N and part-C respectively hereafter) of the enzyme on our complex crystal structure served as a relevant structural model (Fig. 4a) 13,17 . Even though part-C and part-N have been structurally characterised separately, they do not appear to be distinct functional units. Rather, intimate contacts between part-N and -C are present, even in the loop spanning amino acids, which links these two units. Also, ATasefragments lacking part-C or portions thereof do not have AMPylation activity and do not cause AMPylation-mediated cellular toxicity. Instead, part-C interacts with helix α3 and the loop α3-β2, which is localised in the catalytic centre of the ATase loop ( Supplementary Fig. 12), suggesting that these interactions are necessary for correct folding of this region and positioning of catalytically relevant amino acids. This view is supported by the fact that most interactions (nine out of ten amino acids) between DrrA 16-352 with the Rab-protein are formed by part-N (Supplementary Fig. 12 and Fig. 2c). The only amino acid interacting with the Rab from part-C is E264 DrrA and its mutation to alanine impairs the AMPylation activity moderately (Fig. 3b). Therefore, the lack of enzymatic activity of constructs lacking part-C cannot be explained by the deletion of essential amino acids involved in Rab-binding. Instead, this fact supports the notion that part-C stabilises regions in part-N necessary for catalysis.
Albeit no global structural differences between the composite putative apo structure and the full ATase domain occur (Fig. 4b), significant structural differences can be observed for the secondary structure elements in the vicinity of the DrrA ATase active site (Fig. 4c). Indeed, upon Rab binding, helix α3 may be pulled into the protein core, and D110 DrrA (β2), D112 DrrA (β2) and D150 DrrA (α4−α5 loop) move together so that the active Mg 2+ coordinating function of these amino acid residues can be fulfilled. Of note, we previously reported that AMPylationcompetent DrrA 8-533 is not able to bind ATP in the absence of Rab1 25 . Therefore, Rab binding to the NC-RBS of DrrA may lead to rearrangement of the enzyme active site. We therefore wondered whether Rab binding to the NC-RBS may regulate DrrA's AMPylation activity by reordering the catalytic residues. In order to investigate the potential influence of Rab1b binding to the ATase domain of DrrA (i.e. DrrA 16−352 ), we analysed the rate of Rab1b AMPylation in dependence on the absolute concentration of active Rab1b:GppNHp by measuring the decrease in Rab tryptophan fluorescence upon AMP modification. We observed a lag-phase in the AMPylation rate at low Rab1b:GppNHp concentrations, indicative of a stimulatory contribution of Rab1b binding at higher concentrations (Fig. 4d). Therefore, fitting the data to a simple Michaelis-Menten model (i.e. to a hyperbolic function) was not feasible; hence, a sigmoidal Hill-type function was instead employed. The cooperativity parameter resulting from the Hill fit was greater than 1 (n = 1.7 ± 0.16), indicating a major contribution of Rab1b:GppNHp-binding to the rate of AMPylation. DrrA  containing all three domains shows an identical kinetic profile (n = 1.6 ± 0.20) (Fig. 4e).
Furthermore, we also tested the stimulatory function of Rab1b binding to DrrA in cellulo in order to exclude the possibility of in vitro artefacts. Since DrrA causes AMPylation-induced cellular cytotoxicity in mammalian cells upon overexpression, the activity of individual DrrA mutants can be determined by quantifying the cell viability 13 . To this end, eukaryotic H1299 cells were transfected with the previously characterised alanine substitutions (R70A DrrA , Q71A DrrA and K74A DrrA ) of an N-terminally eGFP (enhanced green fluorescent protein)-tagged DrrA 8−533 construct. DrrA 8−533 was employed since the presence of the GEF domain (DrrA 340−533 ) is required to produce active, GTP-loaded Rab1b in the cell. Positively transfected cells were collected by fluorescence-activated cell sorting (FACS) and used for cell viability analysis by the absorbance-based MTS (3-(4,5dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium) assay. DrrA 8−533 but not the AMPylation-deficient D110A/D112A DrrA mutant caused pronounced cytotoxicity 13 . However, H1299 cells transfected with DrrA 8-533 or with these selected single alanine mutant R70A DrrA , Q71A DrrA or K74A DrrA showed similar cytotoxicity (Fig. 4f). Although kinetics indicated that single alanine mutants (R70A DrrA , Q71A DrrA or K74A DrrA ) dramatically decrease the enzyme activity in vitro, results from experiments in cellulo differ from in vitro studies probably due to the longer duration of the in vivo experiments. To make an approximate comparison, we investigated Rab1b-AMPylation by the DrrA single mutants R70A DrrA , Q71A DrrA or K74A DrrA via intact high-resolution mass spectrometry at 72 h incubation time. Extended incubation times of the single mutants resulted in nearly quantitative AMPylation of Rab1b ( Supplementary Fig. 13). Structural analysis of the DrrA:Rab8a-complex indicated that these three residues can interact with each other to constitute a triangular network, which may stabilise the interaction with D44 from Rab1b/Rab8a. We therefore reasoned that the simultaneous combination of these mutations is required to result in a reduction in cytotoxicity compared to wt DrrA  . The triple mutant R70A/Q71A/ K74A DrrA (corresponding to the NC-RBS) of DrrA 8−533 was not cytotoxic (Fig. 4g). These data show that mutations in the NC-RBS affect cellular AMPylation and DrrA-mediated cytotoxicity, indicating that the AMPylation activity of DrrA is regulated via Rab1b:GppNHp binding to the non-catalytic site. To further prove that mutations in the NC-RBS site of DrrA occlude binding of Rab under physiological conditions and thereby affect DrrA activity, we set out to perform our UAA-based, proximitytriggered crosslinking approach in living HEK293T cells. We coexpressed BrC6K-bearing Rab1b (Rab1b-R69BrC6K) together with different non-toxic eGFP-DrrA 8−533 variants in HEK293T cells. Crosslinking was specific for Rab1b-R69BrC6K and the previously identified D82C mutant of eGFP-DrrA 8−533. In accordance with cytotoxicity data, the triple alanine mutant eGFP-DrrA 8−533 (R70A/Q71A/K74A) bearing a D82C mutation was deficient in forming a crosslinked complex with Rab1b-R69BrC6K ( Fig. 4h and Supplementary Fig. 14).
In addition, we wondered whether the presence of the AMP group in Rab1b at Y77 Rab1b would exclude its binding to the NC-RBS. We therefore utilised our established approach and monitored time-resolved crosslink formation (via D82C DrrA ) of either AMPylated or non-AMPylated Rab1b-R69BrC6K-His6 with the DrrA 16-352 -D82C-D110A-D112A variant by SDS-PAGE and western blotting. The AMPylation-deficient D110A/D112A DrrA-mutant was used to exclude modification of non-AMPylated Rab1b during the experiment. Indeed, both AMPylated and non-AMPylated Rab1 form a crosslinked complex in vitro with minor differences in the rate of reaction ( Supplementary Fig. 15). Hence, the presence of AMP in Rab1b does not exclude binding to the NC-RBS.
In summary, biochemical experiments, cytotoxicity data and crosslinking approaches confirm that the AMPylation activity of DrrA is allosterically stimulated by Rab1b-binding to the NC-RBS. However, to what extent the binding of Rab1b to the NC-RBS leads to a structural reorganisation of the catalytic site cannot be deduced entirely from the available structure comparisons: In the absence of the C-terminal part of the ATase-domain (i.e. part-C), essential intramolecular interactions between part-C and part-N are lacking and thus the catalytic Formation of a Rab-DrrA complex via the catalytic site. In addition to assessing the significance of the NC-RBS in DrrA, we attempted to create a complex with the Rab protein bound to the catalytic site of DrrA. For this purpose, we applied the same crosslinking strategy by combining TReND-1 with Cys substitutions located in close proximity to the putative ATP-binding site in DrrA 16−352 . We produced the A176C, D177C and T181C variants of DrrA as potential sites for TReND-1 reaction since these amino acids are likely to be close to the ATP nucleobase interaction site (Fig. 5a). Indeed, DrrA 16−352 containing the A176C substitution, but not D177C or T181C, formed a covalent ternary complex with Rab1b 3−174 in vitro, as indicated by an increase in molecular weight observed by SDS-PAGE (Fig. 5b). Using this approach, we were able to obtain milligram amounts of the ternary DrrA 16−352 -A176C:TReND-1:Rab1b 3−174 complex with high purity, as demonstrated by intact HR-MS (Fig. 5c). Structure determination of the complex failed due to a lack of diffracting crystals. Nevertheless, the ternary complex from the catalytic site of DrrA permitted us to further validate the presence of NC-RBS of DrrA by evaluating and comparing the catalytic activity. For this purpose, we compared the relative AMPylation activity of DrrA 16−352 with the ternary complexes linked via the NC-RBS and the catalytic site using the time-resolved change in tryptophan fluorescence of Rab1b in response to AMPylation (Fig. 5d). As expected, the DrrA:Rab1 complex linked via the NC-RBS stimulated the AMPylation activity compared with free DrrA 16−352 . By contrast, the complex linked via the catalytic site was unable to AMPylate Rab1b, showing that the presence of the covalently linked Rab1b blocks the access of substrate molecules. Therefore, our results imply that DrrA 16−352 contains two separate Rab-binding platforms. Hence, we propose that allosteric binding of active Rab1 to the NC-RBS switches the active site of DrrA ATase from an unstructured AMPylation-deficient state to an organised AMPylation-competent state, which further mediates AMPylation activity (Fig. 5e).

Discussion
Infection of eukaryotic cells by L. pneumophila is a coordinated process characterised by the controlled release and activity regulation of bacterial proteins. The different enzymatic activities of DrrA (GEF and AMPylation) appear to be influenced in particular by the secretion of the protein during the early phases of infection, and the regulated localisation at the LCV 28 . However, in addition to the AMPylation of Rab1 by the N-terminal ATase domain, DrrA can also modify a number of other Rab GTPases, which could cause high cytotoxicity in the case of excessive and mislocalised activity 13 . To avoid these undesirable side effects, which potentially could diminish the chances of an effective infection, additional direct control of the AMPylation activity of DrrA in L. pneumophila may be important. Using two different proximity-triggered and site-specific chemical crosslinking approaches employing UAAs bearing alkylbromide moieties or thiol-reactive ATP-derivatives combined with cysteinesubstituted DrrA-variants, we discovered a non-conventional Rab1-binding site (NC-RBS) in DrrA. Also, the structure of the DrrA ATase permits us to produce a composite structural model by superimposing previous structures 13,17 , in which three different Rab1 binding sites can be visualised (Fig. 5e). The NC-RBS profoundly contributes to the regulation of the AMPylation activity of DrrA in vitro: GTP-loaded Rab1b binds to an allosteric site located opposite to the catalytic centre of the AMPylation domain and stimulates AMP transfer to the GTPase. Thus, two molecules of Rab1b:GTP simultaneously bind to DrrA: one to the allosteric NC-RBS, the other to the catalytic AMPylation site. Hence, DrrA binds to the LCV by virtue of the PI4P binding P4M and catalyses the activation of Rab1 via GDP-to-GTP-exchange using its GEF domain. This displaces Rab1 from GDI and leads to membrane binding through the C-terminally attached geranylgeranyl-lipids. Subsequently, Rab1:GTP could hypothetically stimulate the AMPylation activity of DrrA by binding to the NC-RBS of the ATase-domain, leading to AMPylation of other recruited Rab1 molecules (see model in Fig. 5f). However, this allosteric activation of DrrA has only been shown in vitro and further in vivo work would be necessary to confirm the relevance of this mechanism for infection by L. pneumophila.
Interestingly, DrrA is present at the LCV only during the early stages of infection, indicating that it may be released from the membrane to the cytosol as maturation of the compartment progresses 28 . Consequently, there may be a risk of global AMPylation by DrrA with deleterious cellular side effects that may impair infection progression. In this regard, the binding of Rab1b:GTP to a regulatory site in the enzyme perhaps functions as a safety mechanism that limits rogue DrrA activity. Since activated Rab GTPases are exclusively membrane-localised, cytosolic DrrA may not be activated by the cytosolic pool of GDP-bound Rab1b 29 . However, in the GDP form, Rab-proteins are complexed with GDI in the cytosol, thereby making switch II inaccessible for AMPylation by DrrA. But the Rab1:GDI complex is weakly dynamic because cellular GEFs are known to contact switch II, leading to GDI displacement by loading the Rab with GTP. Thus, it is conceivable that DrrA could also AMPylate GDP-bound Rab whenever it has been spontaneously dissociated from GDI, explaining the need for an additional safety mechanism, i.e. allosteric activity control of DrrA-AMPylation activity by Rab1b:GTP-binding. Nevertheless, the hypothesis remains speculative at this point. There are several instances where Legionella effectors are activated by host proteins. Legionella effector SidJ requires the host protein calmodulin for activation 30,31 . Upon calmodulin binding, SidJ is activated and polyglutamylates SidE (another Legionalla effector) to inhibit the ubiquitin ligase activity of members of the SidE family. This inactivation is required for successful Legionella replication 30,31 . Secondly, Legionella effector VipD lipase requires Rab5:GTP-dependent activation for phospholipase A1 activity-based targeting of enzymes to endosomes, and this eventually inhibits phagosome maturation to ensure the survival of bacteria within the host 32 . Furthermore, there are multiple examples of effector enzymes from other Gram-negative pathogens such as Salmonella, Yersinia and Vibrio that require host protein-dependent activation 33 . After activation by host proteins, these pathogenic effectors cause numerous PTMs such as ADP-ribosylation 34,35 , phosphorylation 36,37 , acetylation 38 and proteolytic cleavage 39,40 . However, AMPylation that is dependent on activation by a host protein has not been reported previously. Therefore, DrrA comprises a unique example that requires activation by its substrate.
The mode of regulation of the AMPylation activity of DrrA is different from other DNA polymerase β-like enzymes and AMP transferases. For example, AMP-transferring GS-ATase is controlled via binding to the regulatory protein PII, and this molecule binds to a linker region connecting the AMP transferase and AMP removase domains of GS-ATase 18 . Complex formation with PII switches on the AMP transferase site of the enzyme, while turning off the hydrolysis activity of the AMP removase 41,42 . The biochemical and structural modes of regulation are clearly different from those of DrrA as GS-ATase is not dependent on the activity of its substrate glutamine synthetase, and does not contain an allosteric site structurally homologous to DrrA. Other AMP transferases are structurally different from DrrA, and are therefore expected to employ different mechanisms of activity regulation. For example, Fic family enzymes, which are known to AMPylate diverse targets, generally harbour an inhibitory helix that safeguards their activity 43 . However, controlling the positioning or displacement of this inhibitory motif has not yet been demonstrated. Furthermore, in some instances, pseudokinases can possess AMP transferase activity, as exemplified by the highly conserved enzyme SelO 44 . This enzyme is controlled by intramolecular disulfide bond formation, but no regulatory binding partner has yet been identified. Thus, the regulation of DrrA by its substrate is unique among AMP transferases.
In summary, we used complementary approaches for the production of low-affinity DrrA ATase :Rab complexes using covalent stabilisation by proximity-enabled reactions. Structural investigation and biochemical characterisation revealed a previously unrecognised regulatory site in DrrA ATase that stimulates AMP transfer via allosteric binding of the protein substrate (i.e. Rab1b:GTP). We speculate that the stimulation of DrrA ATase by Rab1b:GTP confines AMPylation to the site of DrrA localisation and Rab activation (i.e. the LCV), thereby limiting the potential cytotoxicity of the Legionella enzyme. It will be interesting to see whether this mechanism can also be demonstrated in vivo during infection by L. pneumophila.
as the template, all the other plasmids are generated by Q5 Site-directed mutagenesis with corresponding primer pairs (New England Biolabs). Rab protein plasmids were cloned before (Supplementary Table 5) 13,47 .
Cloning of DrrA constructs for MTS experiments in H1299 cells. Plasmid pAC-GFP-DrrA  was first prepared with Gibson assembly (New England Biolabs). The primer pair H1 & H2 was used for amplifying the empty pAC plasmid. The DrrA    Table 6).
Protein expression and purification. In general, DrrA constructs are fused to an N-terminal eGFP, which bears an additional 10× histidine tag. In order to further purify DrrA constructs, a TEV cleavage site was introduced for removing the Nterminal eGFP. Also, to facilitate small-scale purification, another Twin Strep tag was inserted between the TEV site and DrrA (His-GFP-TS-TEV-DrrA). Nevertheless, DrrA proteins were expressed using the E. coli Rosetta2 (DE3) strain under standard conditions. Briefly, when the OD 600 of the culture reached 0.6 to 0.8, 1 mM isopropyl βd-1-thiogalactopyranoside (IPTG) was added to induce protein expression and the temperature was switched from 37 to 20°C and maintained overnight for expression in Lysogeny broth (LB) medium. DrrA proteins containing Twin Strep tags were purified with Strep-Tactin Magnetic Microbeads (IBA Lifesciences, Goettingen, Germany) by following the manufacturer's protocol. Twin Strep-tag-free DrrA proteins were purified using Ni-NTA resin. TEV protease was then added to cleave the GFP tag, and proteins were further purified by size exclusion chromatography on a 16/600 Superdex 75 pg column pre-equilibrated with 20 mM HEPES, 100 mM NaCl and 1 mM Tris(2-carboxyethyl)phosphine (TCEP).
Rab8a and Rab1b were prepared as described previously 13,48 . Briefly, Rab1b was fused to a N-terminal His 6 -MBP tag, while Rab8a was fused to a N-terminal His 6 . Rab1b and Rab8a were expressed in E. coli BL21(DE3). When the OD 600 of the culture reached 0.6 to 0.8, 1 mM IPTG was added to induce protein expression and the temperature was switched from 37°C to 20°C and maintained overnight for expression in LB medium. Additionally, Rab8a was co-expressed with the chaperone GroEL/S, and the expression of the GroEL/S chaperone was induced by 1 mg/mL arabinose when OD 600 was 0.4 to 0.5. After cell disruption using a French press, the lysates were cleared by centrifugation (51,428 g, 40 min, 4°C), applied to a Ni-NTA resin (buffer A: 50 mM HEPES-NaOH pH 8, 500 mM NaCl, 1 mM TCEP, 1 mM MgCl 2 , 1 µM GDP) and proteins were eluted with an imidazole gradient (buffer B: buffer A supplemented with 500 mM imidazole). During dialysis with buffer A, the His 6 -tag was removed by TEV digestion. After reverse Ni-NTA chromatography, Rab proteins were further purified by size exclusion chromatography (buffer: 20 mM HEPES-NaOH pH 8, 100 mM NaCl, 1 mM TCEP, 1 mM MgCl 2 , 1 µM GDP). Proteins were at least 90% pure. After concentrating to 10 mg/mL, proteins were stored at −80°C. Nucleotide exchange for Rab1 and Rab8a was performed by following previous protocols. Briefly, 5 mM Ethylenediaminetetraacetic acid (EDTA) was added to Rab1/8a (20 mM HEPES-NaOH pH 8, 100 mM NaCl, 1 mM TCEP, 1 mM MgCl 2 , 1 µM GDP). A 20-fold excess of GppNHp was then added and incubated at 4°C overnight. For Rab8a nucleotide exchange assays, an additional 5% (v/v) glycerol was also added for stabilisation, and the buffer was exchanged using NAP columns (GE Healthcare, Munich, Bavaria, Germany). Nucleotide binding to Rab was confirmed by HPLC with a C18 reversed-phase liquid chromatography column.
The cell pellets were resuspended in 30 mL His-wash buffer containing 50 mM HEPES pH 8.0, 500 mM LiCl, 20 mM imidazole, 2 mM β-mercaptoethanol, 1 mM MgCl 2 supplemented with 0.1 mg/mL DNase I (AppliChem) and two cOmplete TM Mini EDTA-free protease inhibitor cocktail tablet (Roche). For Rab1b proteins the His-wash buffer further contained 0.01 mM GTP. The cell suspension was incubated on ice for 30 min and sonicated with cooling using an ice-water bath.
The cell lysate was centrifuged (24,000 × g, 1 h, 4°C) and the cleared lysate was transferred to 1 mL Ni-NTA slurry (Jena Bioscience) and incubated for 1 h at 4°C slightly shaking. The beads were transferred to a gravity flow column and washed with 200 mL of His-wash buffer before eluting the proteins in 1 mL fractions using His-wash buffer supplemented with 500 mM imidazole pH 8.0. The fractions containing the protein were pooled, concentrated and rebuffered in storage buffer with Amicon ® Ultra-4 10 K NMWL centrifugal filter units (Millipore). Rab1bstorage buffer contained 20 mM HEPES pH 8.0, 150 mM NaCl, 5 mM MgCl 2 , 2 mM DTT and 0.01 mM GTP. DrrA-storage buffer contained 20 mM HEPES pH 8.0, 150 mM NaCl, 1 mM MgCl 2 , and 2 mM DTT. Purified proteins were analysed by 15 % SDS-PAGE and/or mass spectrometry and stored at −80°C. Protein concentration was determined using NanoPhotometer ® N60 (Implen GmbH).
The obtained cell pellets were thoroughly resuspended in 30 mL of Strep-wash buffer containing 100 mM Tris pH 8.0, 150 mM NaCl, 1 mM EDTA and supplemented with 0.1 mg/mL DNase I (AppliChem) and two cOmplete TM Mini EDTA-free protease inhibitor cocktail tablets (Roche). The cell suspension was incubated on ice for 30 min and sonicated with cooling in an ice-water bath. The lysed cells were centrifuged (24,000 × g, 45 min, 4°C), the cleared lysate added to 500 µL of Step-Tactin ® XT Superflow 50 % Suspension (IBA Lifesciences) and the mixture was incubated with agitation for 5 min at 4°C. After incubation, the Step-Tactin ® XT beads were transferred to a gravity flow column and washed with 200 mL of Strep-wash buffer. The proteins were eluted in 1 mL fractions with Strep-wash buffer supplemented with 50 mM biotin. The fractions containing the protein were pooled together, concentrated and rebuffered (20 mM HEPES pH 8.0, 150 mM NaCl, 2 mM DTT, and 5 mM MgCl 2 ) using Amicon ® Ultra-4 10 K NMWL centrifugal filter units (Millipore). Purified proteins were analysed by 15 % SDS-PAGE and stored at −80°C. Protein concentration was determined using NanoPhotometer ® N60 (Implen GmbH).
For preparative expression of crosslinked complex, expression was performed as described above in 1 L AI medium 49 . Purification of in vivo crosslinked Rab1b-DrrA complexes involved a two-step procedure comprising Ni 2+ -affinity and sizeexclusion chromatography. Cell pellets were resuspended in 30 mL His-wash buffer containing 50 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) pH 8.0, 500 mM LiCl, 20 mM imidazole, 2 mM β-mercaptoethanol, and 1 mM MgCl 2 supplemented with 0.1 mg/mL DNase I (AppliChem, Darmstadt, Germany) and two cOmplete Mini EDTA-free protease inhibitor cocktail tablets (Roche). The cell suspension was incubated on ice for 30 min and sonicated with cooling using an ice-water bath. The cell lysate was centrifuged (24,000 × g, 1 h, 4°C) and the cleared lysate was mixed with 1 mL Ni-NTA slurry (Jena Bioscience) and incubated for 1 h at 4°C with gentle shaking. The mixture was poured into a plastic column and washed with 200 mL of His-wash buffer before eluting bound proteins in 1 mL fractions using His-wash buffer supplemented with 500 mM imidazole (pH 8.0). was added and the digest was incubated at 37°C overnight with shaking at 450 rpm. The reaction was stopped by addition of formic acid (more formic acid was added to ensure the pH remained 3 or below). For desalting, stage-tipping was performed with a double C 18 membrane using Octadecyl C18 47 mm Extraction disks (Empore Products, CDS Analytical, Oxford, USA). The membrane was washed three times with 70 µL MeOH, 70 µL buffer E (80% (v/v) acetonitrile, 0.5% (v/v) formic acid), and 70 µL 0.5% (v/v) formic acid before the digested protein sample was loaded. The membrane was washed three more times with 70 µL 0.5% (v/v) formic acid before eluting with twice with 30 µL buffer E into a LoBind tube. Samples were lyophilised using a SpeedVac and stored at −80°C until measurement. Before applying samples to the mass spectrometer, they were dissolved in 30 µL 1% (v/v) formic acid and filtered using Ultrafree-MC-GV centrifugal filters. The filter was washed with 300 µL 1% (v/v) formic acid by centrifugation at 16,000 × g for 2 min before the dissolved sample was filtered (16,000 × g; 2 min) and collected into a new LoBind tube. MS analysis of digested protein samples was performed on an Orbitrap Fusion instrument coupled to an Ultimate 3000 Nano-high-pressure liquid chromatography (HPLC) platform via an electrospray easy source (all Thermo Fisher Scientific). Samples (7 µL) were loaded onto a 2 cm PepMap RSLC C18 trap column (2 μm particles, 100 A, inner diameter 75 μm; Thermo Fisher Scientific) with 0.1% trifluoroacetic acid and separated on a 50 cm PepMap RSLC C18 column (2 μm particles, 100 A, inner diameter 75 μm; Thermo Fisher Scientific) at a constant temperature of 50°C. The gradient was 5-32% acetonitrile, 0.1% formic acid (7 min 5%, 105 min to 22%, 10 min to 32%, 10 min to 90%, 10 min wash at 90%, 10 min equilibration at 5%) at a flow rate of 300 nL/min. Survey scans (m/z 300-1500) were acquired in the Orbitrap with a resolution of 120,000, and a maximum injection time of 50 ms, with an automatic gain control (AGC) target of 4e5. Most intense ions with charge states of 4-8 and an intensity threshold of 5e3 were selected for fragmentation by high-energy collisional dissociation (HCD) with a collision energy of 30%. Fragment spectra were again recorded in the Orbitrap with a resolution of 30,000, a maximum injection time of 100 ms, and an AGC target of 5e4. The 'inject ions for all available parallelizable time' option was enabled. Dynamic exclusion was employed with an exclusion duration of 120 s. The overall cycle time was 5 s.
Data analysis and crosslink detection. Raw mass data files were converted to mzML files using MSConvert in ProteoWizard 19,50 . Crosslink searches were then performed with Kojak software version 1.5.5 (http://www.kojak-ms.org) 51 against a database consisting of Rab1b_Q67A_R69K-DrrA 16-352 , common contaminant proteins downloaded from the Andromeda configuration in the MaxQuant software package 52 , and all reverse sequences. The MS1 and MS2 resolution was set to 120,000 and 30,000, respectively. Variable modifications included oxidation on methionine (+15.9949) and AMPylation on tyrosine (+329.0525). Static modifications included carbamidomethylation (+57.02146) on cysteine. Three modifications were allowed per peptide. The fragment bin offset was set to 0 and the size was 0.03. A maximum of three missed cleavages were allowed. The minimum peptide mass was set to 300 Da. The precursor mass tolerance was fixed at 10 ppm and the settings for the fragment tolerance were set to fit high-resolution MS2 data (fragment_bin_offset: 0.0 Th, fragment_bin_size: 0.03 Th). The unnatural amino acid was encoded as lysine in our database, hence we searched for crosslinks (+96.0575) between lysines and cysteines, serines, threonines, aspartates, glutamates, and protein N-termini. Additionally, we searched for monolinked species (+175.9837). Annotated spectra were visualised by Kojak Spectrum viewer 53 . Further assessment including hit selection and statistical analysis was performed using custom R scripts (https://github.com/higsch/crosslinkR). Therefore, the Kojak output files *.perc.inter.txt were loaded and each putative crosslink was evaluated whether it contained Rab1b and DrrA, and contained no decoy protein at all. Crosslinks fulfilling these criteria are hits (Supplementary Table 1). The script was further used to assess the confidence of the crosslink identifications by setting the hit scores into the context of the score distributions of all decoy and target hits ( Supplementary Fig. 2). Please refer to Supplementary Note 1 on the analysis of mass-spectrometric data to get an in-depth explanation of the underlying statistics and our considerations not using an FDR approach here 51,53 .
MS proteomics data, Fasta files, and Kojak configuration files have been deposited at the ProteomeXchange Consortium (http://proteomecentral. proteomexchange.org.) via the PRIDE partner repository 54 under dataset identifier PXD019043.
Cytotoxicity analysis of DrrA mutants. In order to evaluate the cytotoxicity of different DrrA mutants, lipofectamine LTX was used for transient transfection of H1299 cells with eGFP, wt DrrA, and various other constructs. At 4 h after transfection, the culture medium was replaced with 10% foetal bovine serum (FBS) supplemented Dulbecco's modified Eagle's medium (DMEM). At 24 h after transfection, GFP-positive cells were sorted in the FACS facility of UKE (using a FACS Aria Fusion, see Supplementary Fig. 16 for gating strategy). Next, 24,000 cells were seeded into three wells of a 96-well plate. After 24 h, MTS assays were conducted for cytotoxicity analysis. Cell viability was determined by calculating the ratio of the indicated sample relative to the eGFP vector control. Data were analysed using Origin 2019b (Origin software Inc, Northampton, MA 01060, USA) and Graphpad Prism v 5.0 (GraphPad Software, San Diego, CA 92108, USA).
Structure determination of the DrrA-Rab8a complex. The DrrA 16−352 -L197C: TReND-1:Rab8a 6-176 complex in 20 mM HEPES pH 8, 100 mM NaCl, 1 mM (TCEP), 1 mM MgCl 2 , and 1 µM GppNHp was concentrated to 14 mg/ml using a centrifugal filter device (Millipore), and centrifuged to remove debris (16,000 × g, 15 min, 4°C) prior to high-throughput crystallisation screening using commercially available screening solutions (NeXtal DWBlock Suites, Qiagen) and robotics (Phoenix, Art Robbins). Complex crystals were obtained by the sitting-drop vapour diffusion method in 0.1 M BICINE and 2.4 M ammonium sulphate pH 9 at 20°C. Prior to cryo-cooling and storage in liquid nitrogen, crystals were dipped into reservoir solution supplemented with 2 M lithium sulphate for cryoprotection. Diffraction data were collected at the European Synchrotron Radiation Facility (ESRF, beamline MASSIF-1) and were processed with XDS 55 . Crystals belonged to the P321 space group and the resolution cut-off of 2.15 Å was chosen according to the correlation coefficient of random half datasets (CC 1/2) at~50% [56][57][58] . The structure was solved using the coordinates of the previously reported structures of the N-terminal domain of DrrA 17−210 (PDB: 3NKU) and Rab8 (PDB: 4LHV) by molecular replacement in PHASER 59 . Following simulated annealing with PHENIX 60,61 , peaks of density for residues 210−348 of DrrA16  , which were not present in the search model, were clearly visible. The model was completed by iterative cycles of manual model building in COOT 62 and restraint and TLS (translation/libration/screw) rigid-body motion refinement in REFMAC5 63 . Structure optimisation was carried out using the PDBredo server 64 . Data processing and structure refinement statistics are included in Supplementary Table 1. All structural figures were prepared with PyMol (Schrödinger). Crystallographic data for the DrrA-Rab8a complex have been deposited in the Protein Data Bank (https://www.ebi.ac.uk/pdbe/) under PDB accession code 6YX5.
Temperature-scanning CD measurements. CD signals were recorded on a Chirascan CD Spectrometer (Applied Photophysics, Leatherhead, Surrey, UK) in a 1 mm cuvette with a bandwidth of 0.5 nm and a response of 0.5 s. Temperature scanning CD measurements were taken at 222 nm at a heating rate of 1 K/min. Data were analysed using Origin 2019b and evaluated by the Boltzmann equation to obtain melting temperatures.
AMPylation kinetics of DrrA. The kinetics of Rab1b:GppNHp AMPylation by DrrA was monitored via the change in intrinsic Rab1b tryptophan fluorescence using an F-2710 fluorescence spectrophotometer (Hitachi, Schaumburg, IL, USA) (excitation wavelength of 297 nm, emission wavelength of 370 nm, excitation slit width of 2.5 nm, an emission slit width of 5 nm). The Rab1b concentration was <64 µM. However, for Rab1:GppNHp concentration >64 µM, the emission wavelength was shifted to 390 nm, while keeping other parameters constant. All measurements were conducted in the presence of 100 nM DrrA and 1 unit of pyrophosphate (New England Biolabs) at 25°C with GppNHp in buffer (20 mM HEPES, 100 mM NaCl, 1 mM MgCl 2 , 1 µM GppNHp, 1 mM TCEP, pH 7.5) (Supplementary Fig. 17). Data evaluation was performed as previously described 14 . Initial velocities of Rab1b-AMPylation were obtained by fitting the AMP-Rab1b versus time to equation 1: where v is the initial velocity, c AMP-Rab1b (t) indicates the concentration of AMP-Rab1b at time t and t is time. For yielding the initial velocity (k obs ), the initial velocity v was divided by the enzyme concentration. Hyperbolic curve fitting was performed according to equation 2: where S is the initial concentration of Rab1b, k cat is the turnover number and K M is the Michaelis constant. Sigmoidal curve fitting was performed according to equation 3: where n is the cooperativity parameter.
To determine the AMPylation rates of different DrrA constructs, timeresolved tryptophan fluorescence was applied. In general, 200 µM GTP and 100 nM DrrA GEF were added for the nucleotide exchange of Rab1 (5 µM) from GDP to GTP, and AMPylation was initiated by addition of different DrrA constructs (100 nM) (Supplementary Fig. 17). Data evaluation was performed as previously described 14 . In brief, for the determination of catalytic efficiencies (k cat /K M ) of AMPylation reactions measured by fluorescence spectrometry, reaction curves were fitted to a single exponential curve according to using equation 4: where F(t) is the fluorescence intensity, F 0 is the minimum fluorescence intensity, F A is the total fluorescence amplitude (i.e., F max− F 0 , with F max as the maximum fluorescence intensity), and k obs is the observed rate constant. The observed rate constant (k obs ) was divided by the applied DrrA concentration (100 nM), yielding k cat /K M .
Reporting Summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
Atomic coordinates and structure factors of the DrrA 16-352 :TReND-1:Rab8a 6-176 complex have been deposited in the Protein Data Bank with accession code 6YX5. The mass spectrometry proteomics data have been deposited at the ProteomeXchange Consortium (http://proteomecentral.proteomexchange.org/) via the PRIDE partner repository, with data set identifier PXD019043. The data that support the findings of this study are available from the corresponding author upon reasonable request. Source data are provided with this paper.