Structural Insights into a Unique Legionella pneumophila Effector LidA Recognizing Both GDP and GTP Bound Rab1 in Their Active State

The intracellular pathogen Legionella pneumophila hijacks the endoplasmic reticulum (ER)-derived vesicles to create an organelle designated Legionella-containing vacuole (LCV) required for bacterial replication. Maturation of the LCV involved acquisition of Rab1, which is mediated by the bacterial effector protein SidM/DrrA. SidM/DrrA is a bifunctional enzyme having the activity of both Rab1-specific GDP dissociation inhibitor (GDI) displacement factor (GDF) and guanine nucleotide exchange factor (GEF). LidA, another Rab1-interacting bacterial effector protein, was reported to promote SidM/DrrA-mediated recruitment of Rab1 to the LCV as well. Here we report the crystal structures of LidA complexes with GDP- and GTP-bound Rab1 respectively. Structural comparison revealed that GDP-Rab1 bound by LidA exhibits an active and nearly identical conformation with that of GTP-Rab1, suggesting that LidA can disrupt the switch function of Rab1 and render it persistently active. As with GTP, LidA maintains GDP-Rab1 in the active conformation through interaction with its two conserved switch regions. Consistent with the structural observations, biochemical assays showed that LidA binds to GDP- and GTP-Rab1 equally well with an affinity approximately 7.5 nM. We propose that the tight interaction with Rab1 allows LidA to facilitate SidM/DrrA-catalyzed release of Rab1 from GDIs. Taken together, our results support a unique mechanism by which a bacterial effector protein regulates Rab1 recycling.

The intracellular bacterial pathogen Legionella pneumophila (L. pneumophila) is the causative agent of pneumonia Legionnaires disease [8,28]. Following invasion of host cells, L. pneumophila resides in the LCV [29] that escapes endolysosomal destruction [30]. The bacterial effector proteins, delivered by the type IV secretion system (T4SS) of L. pneumophila into the cytosol of host cells [31], actively remodel the LCV to establish an intracellular niche indispensable to bacterial pathogenesis [32]. For example, the early secretory vesicles from ER can be hijacked to the LCV, converting it into an ER-derived organelle that supports bacterial replication [33][34][35][36]. Rab1, required for vesicle trafficking between ER and the Golgi complex [18,37,38], is one of the host proteins recruited to the LCV shortly after uptake of L. pneumophila [39][40][41]. Rab1 recruitment subsequently promotes transport and fusion of ER-derived vesicles with the LCV, thus playing an essential role in the biogenesis of this organelle [42,43]. The effector protein SidM/DrrA, by acting as a Rab1-sepcific GEF and GDF, is required for LCV recruitment of Rab1 [44,45]. Further, the AMPylation activity of SidM/DrrA modified Rab1 by covalently adding adenosine monophosphate (AMP) to avoid the GAP recognition [46]. SidD, functions as the Rab1 deAMPylase, generating de-AMPylated Rab1 accessible for inactivation by LepB [43,47]. Another translocated effector protein LidA is also involved in the recruitment of early secretory vesicles to the LCV [48,49]. LidA was found to interact with Rab1 as well, regardless of nucleotide binding states, and promote SidM/DrrA activity of transporting Rab1 to the LCV [39,45]. Unlike SidM/DrrA mutants, however, L. pneumophila mutants lacking LidA displayed a temporal delay but not a loss of Rab1 recruitment to the LCV [39]. Intriguingly, while indispensable to the recruitment of wild type Rab1, SidM/DrrA is not required to accumulate Rab1 mutant (D44N) that loses interaction with GDIs but not with LidA about the LCV [45]. These results suggest that GDIs play a negative role in Rab1 recruitment by the LCV during L. pneumophila infection. Recruitment of this Rab1 mutant, however, is dependent on LidA, suggesting that interaction with Rab1 is important for the role of LidA in delivering Rab1 to the LCV. Currently the mechanisms of how LidA cooperates with SidM/ DrrA for Rab1 recruitment are not well understood.
Here, we present the crystal structures of LidA (residues 224-559) in complex with a GDP-bound Rab1 mutant (S25N; residues 1-176) [50,51] and LidA (residues 188-449) in complex with GTPbound Rab1 (residues 1-176). Unexpectedly, the structures showed that GDP-Rab1(S25N), a ''constitutively'' inactive mutant, adopted an active conformation when bound by LidA. In agreement with the structural observation, biochemical assays demonstrated that GDP-Rab1(S25N) and GTP-Rab1(Q70L) [52] exhibited a similar and exceptionally high binding affinity for LidA. Coupled with previous observations, data presented in current study support a unique mechanism by which LidA interferes with the host secretory vesicular trafficking.

LidA binds both GDP and GTP bound Rab1 with a similar affinity
The Rab1-interacting domain of LidA used in our study is similar to the one (residues 191-600) necessary and sufficient to disrupt the secretory pathway when overexpressed in COS1 cells [49]. Consistently, In vitro data shows truncation of 188 residues from the N-terminal side or 155 residues from the C-terminal side of LidA generated no effect on the formation of such a complex ( Figure S1). To further validate the interaction between LidA and Rab1, we measured their binding affinity using isothermal titration calorimetry (ITC) technique. The ITC results showed that the LidA fragment 188-580 containing the predicted coiled-coil domain interacted with the constitutively active mutant Rab1(Q70L) with a dissociation constant of 7.5 nM ( Figure 1A). Surprisingly, the same protein also bound nearly equally well to the constitutively GDP-bound mutant of Rab1(S25N), with a dissociation constant of 7.6 nM ( Figure 1B). A preference of LidA for the GTP-Rab1 could result from a lower abundance of GDP-Rab1 in cells.

Overall structures of LidA and its complex with Rab1
To reveal the molecular basis for the LidA-Rab1 interaction, we first solved the crystal structure of LidA(224-559) in complex with a constitutively GDP-bound mutant Rab1(S25N; 1-176) at 1.73 Å using molecular replacement (Table S1 in supporting information).
The overall LidA-Rab1 complex adopts a compact and globular structure (Figures 2A,B). The interaction between LidA and Rab1 resulted in a 1:1 stoichiometric complex and buried 41% (4031 Å 2 /9845 Å 2 ) of the Rab1. The exceptionally large buried surface generated by LidA-Rab1 interaction is consistent with their strong binding affinity. In the structure, Rab1 is held in a large and pronounced groove made by the four ''fingers'' with an extensive, though not complete, charge and surface complementarities. Contacts between LidA and Rab1 are established through a combination of hydrogen bonds and hydrophobic interactions

Author Summary
Legionella pneumophila delivers 275 validated substrates into the host cytosol by its Dot/Icm type IV secretion system. Several substrates including SidM/DrrA and LidA directly interact with the host Rab GTPases and interfere with the vesicle secretion pathway. SidM/DrrA is necessary for Rab1 recruitment, function as a Rab1 specific GDI displacement factor and guanine nucleotide exchange factor. LidA has the auxiliary activity for Rab1 recruitment, whereas it is more important for the formation of the replication vacuole compared with SidM/DrrA. LidA is predicted to be the first substrate secreted by the Dot/Icm system and is critical for maintaining the integrity of the bacterial cell. Moreover, it expresses throughout the intracellular growth phase, localizes to early secretory compartments, and interacts with several members of Rab family. Here we present the crystal structures of LidA coiled-coil domain in complex with two different states of Rab1, GDP-and GTP-bound. The GDP-bound Rab1 in the complex surprisingly has the same conformation with the GTP-bound Rab1, revealing that LidA can retain Rab1 persistently in its active state. Our structures add a new insight into the regulation of the host Rab1 membrane cycle by pathogen-secreted coiled-coil effector.
involving the switch I, switch II and the interswitch region of Rab1 ( Figure 2B).
LidA (residues 224-559) is composed of seven a-helices and ten b-strands, which arrange into a structure resembling a hand with the baby finger buried in the palm and the remaining four straightened ( Figure 2C). Six of the seven a-helices form three extended two-stranded a-helical coiled-coils. This structural observation supports a previous prediction that LidA is a coiledcoil rich structure [48,49]. Two of them, made by a4/a5 and a1/ a7, are the ''index finger'' and the ''middle finger'', respectively. Interaction between them is primarily mediated by packing of a1 against a5. The ''thumb'' stems from a two-helix coiled-coil formed by a2 and a3 helices which interact with the base of ''index finger'' through packing against a4. The isolated helix a6 and its following loop make extensive hydrophobic contacts with three anti-parallel b-sheets (b5/b6, b7/b8 and b9/b10), resulting in formation of the ''ring finger''. The N-terminal side of a4, which is not involved in the formation of coiled-coil structures, tightly contacts the two anti-parallel b-sheets (b1/b2 and b3/b4) located between a3 and a4, thus making the ''wrist''. Interestingly, the a1 and a7 stay closely together and form the middle finger. The terminal parallel helix of LidA extends to the outside of interacting region, forms an antiparallel long coiled-coil ( Figure 2C). So we suppose the N-and C-terminal domains of LidA may stay spatially close cooperate with each other to perform specific function.

Interaction interfaces between LidA and Rab1
In the complex, switch I (residues 33-48) of Rab1 is mainly sandwiched between the ''middle finger'' and the ''ring finger'' by making contacts with a1, a7 and a6. It also contacts the ''index finger'' a5 ( Figures 3A,B). Close packing of Ile44 Rab1 from switch I and Phe73 Rab1 from switch II (residues 65-83) against Leu541 LidA , Val542 LidA , Val538 LidA , Leu436 LidA , Ala439 LidA , Y243 LidA and the aliphatic portion of Asn432 LidA (from a5) appears to dominate the interactions around this interface. Additional van der Waals interactions result from contacts of Tyr532 LidA with the alpha carbon atom of Gly21 Rab1 and Tyr40 Rab1 , and Ile41 Rab1 with Leu548 LidA and Glu549 LidA in LidA. Hydrogen bonds further strengthen this interface. In particular, in addition to forming a salt bridge with Asp443 LidA , Arg72 Rab1 from the switch II region makes a hydrogen bonds with the backbone carbonyl of Met536 LidA . Through an intramolecular hydrogen bond, Ar-g72 Rab1 is stabilized by the catalytically important residue Gln70 Rab1 , which in turn mediates a bifurcated polar interaction with the carbonyl oxygen of Thr534 LidA and Glu533 LidA from LidA.
The C-terminal portion of the switch II, together with the region defined by b2, b3 and b4, makes extensive contacts with the base of ''index finger'' (Figures 3A,C). Center to this interface are the van der Waals interactions made by Phe48 Rab1 and Trp65 Rab1 from Rab1 with a cluster of surrounding hydrophobic residues of LidA. At periphery, a number of hydrogen bonds flank the hydrophobic interaction center. Among these, Lys61 Rab1 from Rab1 forms three hydrogen bonds with Asn421 LidA and the carbonyl oxygen atoms of Ile413 LidA and Ser415 LidA . Other hydrogen bonds include those formed by Tyr8 Rab1 with the carbonyl oxygen atom of Met414 LidA , Gln63 Rab1 with Asn410 LidA , the carbonyl oxygen atoms of Val46 Rab1 , ILE44 Rab1 and Phe73 Rab1 with Asn432 LidA and Asn435 LidA , respectively. In addition to hydrogen bonding with His431 LidA , Tyr80 Rab1 also engages hydrophobic contact with Leu406 LidA .

The GDP-bound Rab1(S25N) is held in the active conformation upon LidA binding
Our biochemical assay ( Figures 1A,B) indicates that LidA binding is independent of the Rab1 nucleotide-binding states. To understand the underlying molecular mechanisms, we solved the crystal structure of LidA (residues 188-449) in complex with the Rab1(WT; residues 1-191) at 2.2 Å . The structure of the complex is highly similar to that of LidA(224-559)-Rab1(S25N) with an r.m.s.d of 0.678 Å for Rab1; 1.496 Å for LidA, respectively ( Figure 4A). As anticipated, GTP is well defined in the crystal structure ( Figure 4A) and its interactions with Rab1 are conserved in other active Rabs ( Figure S2) [53][54][55]. Surprisingly, structural comparison revealed that the conformation of LidA-bound Rab1(S25N) is essentially identical with that of LidA-bound GTP-Rab1 ( Figure 4B), indicating that Rab1(S25N) is in the active conformation following LidA binding. While the switch II region in the free GDP-Rab1 is completely disordered ( Figure 4C), it is well defined when bound by LidA. As with GTP, LidA binding induces striking structural remodeling around the switch I region of Rab1(S25N), which swings from the edge of b-sheet to the nucleotide-binding site, with the largest displacement of 9.2 Å at the Ca atom of Ile. Relocation of switch I results in formation of an extended b-strand b2 at its N-terminal side and hydrophobic packing against switch II ( Figure 4C).
Our structural analyses suggest that LidA utilizes a similar mechanism to GTP for maintaining Rab1 in the active conformation. The hydrogen bonding interactions between the c phosphate group of GTP with Thr43 and Gly69, from switch I and switch II of Rab1 respectively, are highly conserved among Rab proteins and important for stabilizing their active state ( Figure  S2). Despite loss of the c phosphate group-mediated interactions, the switch I region in Rab1(S25N) is stabilized through its interaction with a7 of LidA ( Figure 4D). Stabilization of the switch II region in Rab1(S25N) is via its polar interaction with the loop linking a6 and a7 as well as hydrophobic contacts with a5 of LidA ( Figure 4E). Deletion of the switch-stabilizing structural elements would be expected to disfavor LidA interaction with the GDPbound Rab1 as compared to the GTP-bound Rab1. We therefore quantified LidA (residues 188-449) interaction with Rab1(S25N) and Rab1(Q70L) using an ITC assay. In support of the structurebased prediction, the assay showed that the LidA fragment exhibited a higher binding affinity (0.43 mM) to the GTP-bound Rab1 than to the GDP-bound Rab1 (2.4 mM) ( Figure 4F).
Interaction of Rab1-LidA shares similar features with that of Rab4-Rabenosyn-5 Interaction of the anti-parallel coiled-coil formed by the helices a4 and a5 with the switch and interswitch regions of Rab1 is reminiscent of the structure of Rab4-Rabenosyn-5 complex [54]. A superposition of the overall structures of these two complexes revealed significant structural similarity ( Figure 5A). Some subtle structural differences surrounding the switch I and II are consistent with the notion that the switch regions adopt specific conformations in the GTP-binding form of Rab proteins [55]. Rabenosyn-5 forms a similar anti-parallel coiled-coil to LidA and binds the switch and interswitch regions of Rab4. Interestingly, although these two peptides have reverse orientations, they contact a significant subset of the highly conserved resides between Rab1 and Rab4, though the detailed molecular interactions differ. For example, in both complexes the highly conserved residues Phe48 Rab1 (Phe45 Rab4 ) and Trp65 Rab1 (Trp62 Rab4 ) are buried in a cluster of hydrophobic residues despite their different side chain rotamer conformations, whereas Lys61 Rab1 (Lys58 Rab4 ) and Tyr80 Rab1 (Tyr77 Rab4 ) are involved in polar interactions ( Figures 5B,C).

LidA interacts with multiple Rabs
In addition to Rab1, LidA has been shown to be able to interact with other Rab member proteins such as Rab6a and Rab8b [39]. Indeed, many LidA-interacting residues in Rab1 are highly conserved in these two Rabs as well as other Rab members ( Figure 6A) suggesting that LidA may have promiscuity of binding Rab proteins. To further experimentally test this, we purified 15 GST-fused Rab family members (His-fused Rab20 is exceptional) and tested their interaction with LidA using pull down assay. As shown in Figure 6B, thirteen out of the fifteen Rabs tested interacted with LidA. The binding affinities of LidA for other Rabs, including Rab2, Rab4, Rab6, Rab7, Rab9, Rab11, Rab20 and Rab22 have been quantified by ITC. All give K D values in the micromolar range. Compared with them Rab1 gives the strongest binding affinity (The K D is in nanomolar range). Thus Rab1 should be the preferred substrate of LidA compared with other Rabs (Figure 7). Thus, although previous LCV analysis identified interacting Rab GTPases such as Rab1, Rab7, Rab8, and Rab14 as novel LCV components [56], the biological relevance of the interactions await further investigation. To make clear how LidA choose substrate from so many Rabs, we attempted to disrupt the interactions of Rab1 and LidA by making single or double mutations of Rab1 and LidA along the surface of the interactions. However, none of the mutants we have tested by the pull-down assay lose their interaction with LidA. We also tested such mutants in other Rabs and did not observe an effect. So it remains undetermined how LidA selects different Rabs. Nonetheless, it is expected that the mechanism of Rab1 recognition by LidA is applicable to other Rab members given the high conservation in the LidAinteracting surface.

Discussion
The crystal structures presented in current study revealed that the GDP-bound Rab1 is held in the active conformation through associating with LidA. Consistent with this structural observation, quantification assay using ITC demonstrated that LidA binds strongly to both GDP-and GTP-Rab1 with a nearly equal affinity ( Figures 1A, B). These results indicate that Rab1 recognition by LidA is independent of its cycling between GDP-and GTP-bound states. To LidA, the switch mechanisms of Rab1 are not operational and thus Rab1 is rendered persistently active in interaction with LidA. Such a unique function of LidA provides explanations for a body of previous observations.
The prenylated GDI-free Rab1(D44N) was shown to be primarily membrane-localized, indicating that GDI is not absolutely required for membrane targeting of Rab1 [57]. This principle is likely applicable to the delivery of Rab1(D44N) to the LCV [45]. The strong interaction with LidA can be important for the membrane retention of Rab1(D44N), as LidA-deficient mutants displayed a scattered distribution of this Rab1 mutant in COS-1 cells [45]. Because no GDP-Rab1(D44N)-GDI complex is formed and LidA is able to recognize GDP-Rab1, GDF and GEF activity would be made redundant under these conditions, providing an explanation why recruitment of Rab1(D44N) to LCV is independent of SidM/DrrA [45]. To wild type Rab1, the complex of GDP-Rab1-GDI is formed with a high affinity. The direct competition of LidA with GDI for binding GDP-Rab1 will be difficult and inefficient. Thus GDP-Rab1 has to be released from GDI by SidM/DrrA for LidA recognition [39]. These arguments likely hold true with the recruitment of the GDPrestricted Rab1 mutant (S25N) to the LCV. In uninfected cells, the dominant inhibitory effect of Rab1(S25N) derives from the competition of this Rab1 mutant with wild type Rab1 for binding GEFs, resulting in formation of ''dead-end'' Rab1-GEF complexes [58]. In the case of wild type Rab1 recruitment to the LCV, this situation could be altered, because interaction of GDP-bound Rab1-SidM/DrrA complex is transient and much weaker than that of GDP-bound Rab1-LidA complex [39,45] (Figure S3). LidA is therefore expected to be able to outcompete with SidM/DrrA for binding GDP-Rab1, which in turn facilitates the release of GDP-Rab1 from GDF (see further discussion below). As GDP-Rab1 is in the active conformation when complexed with LidA, it is conceivable that Rab1(S25N) can still support the functions performed by wild type Rab1. In complete agreement with this hypothesis, Rab1 binding to the LCV precedes the delivery of Sec22b, GDP-locked Rab1(S25N) can delay but not block the recruitment of Sec22b. This Rab1 mutant can support growth of L. pneumophila, at least for the first 10 hours of postinfection [42].
The strong association with LidA can facilitate the recruitment of Rab1 to LCV by SidM/DrrA. Release of GDP-Rab1 from GDIs catalyzed by SidM/DrrA (as a GDF) is a dynamic process and GDP-Rab1-SidM/DrrA is an intermediate of the reaction [59][60][61]. LidA is expected to displace SidM/DrrA from the intermediate because of its much higher binding affinity to GDP-Rab1, thus shifting the equilibrium toward dissociation of GDP-Rab1-GDI complex. In this respect, LidA is similar to GTP in that the latter is able to exchange GDP from the intermediate and form the GTP-bound Rab1, a product that is unfavorable for interaction with GDFs and thereby promotes disruption of GDP-Rab1-GDI complex. It can be imagined that lack of LidA would slow down the release of GDP-Rab1 from GDI. Consistently, deletion of LidA resulted in delayed but not a blocked recruitment of Rab1 to the LCV [39]. The inability of LidA to discriminate between GDP-and GTP-bound Rab1 may also be advantageous for L. pneumophila to retain the recruited Rab1 on the LCV. Because even GTP-bound Rab1 is subjected to hydrolysis by GAPs on LCV [43,44], the resulting product of GDP-Rab1 is still able to be captured by LidA with a strong binding affinity. Moreover, interaction with LidA would make GDP-Rab1 less susceptible to extraction from the membrane of LCV by GDIs, because they have to outcompete LidA for Rab1 binding.
Unlike other small GTPase effector proteins which mainly binding to Rab1 through interacting with the switch I and II regions that determine nucleotide-dependent interaction, LidA holds Rab1 tightly in its hand like structure with four fingers. It is unlikely that any Rab1 effector can displace LidA from Rab1-LidA complex. LidA has been known to associate with the LCVs for much longer time than SidM/DrrA. However, Rab1 does eventually leave the LCV [44,48], so that a mechanism is needed to decrease the binding affinity between Rab1 and LidA. Since Rab1 is deeply buried in the complex, the most efficient way to release Rab1 is to force LidA to open its fingers by downstream proteins. The GTP-Rab1-LidA structure shows truncation of the middle and the ring fingers does not disrupt Rab1-LidA complex, so that the thumb or the index fingers, though necessary, may not be sufficient to release Rab1 ( Figure 4A). Structure analysis indicates that the thumb and the ring finger do not pack as tightly with other part of LidA as the index and middle finger. This means that LidA's wrist (the N-terminal half of a4 and b1-b4) which is isolated from the central interacting region, the bottom of the ring finger and thumb are accessible for other proteins to trigger a conformational change that releases Rab1 ( Figure S4). Opening of the ring finger could be very important since it can expose the switch region immediately to facilitate Rab1 binding by other effectors.
Recent research showed that the post-translationally modified Rab1b by the L. pneumophila effector protein SidM/DrrA retains the ability to interact with LidA and can avoid the GAP recognition [46]. Our data confirmed that the correspondent residue Tyr80 in Rab1a was also AMPylated ( Figure S5A). Tyr80 is located on the interface between Rab1a and LidA, involved in  (Figure 3C), so it is reasonable to speculate that the covalent-bound AMP group on Tyr80 will affect the stability of LidA-Rab1 complex. However, Rab1 interacts with the LidA palm tightly via an extensive surface, so that the interaction is difficult to interrupt. In the structure of GTP-Rab1-LidA we can see even the middle and the ring fingers are truncated, LidA can still holds Rab1 ( Figure 4A). Moreover the structural feature of four fingers possesses a certain degree of flexibility. Thus the orientations of the four fingers could adjust a little bit to accommodate the additional AMP group. Consistent with previous reports [46] our ITC results showed that the binding affinity of Rab1a(K62H, 1-176) and LidA(188-580) is not significantly affected by the AMPylation (Figure S5B,C). Thus we presume that LidA can form both the Rab1-LidA and AMP-Rab1-LidA complex in vivo. The ability of LidA to form stable complex with AMPylated Rab1 is consistent with its ability to bind a number of different Rabs, which small differences in their surface residues. L. pneumophila may also benefit from this property. Formation of AMPylated-Rab1-LidA complex may protect Rab1 from the de-AMPylation activity of SidD until Rab1 is released from LidA by still unknown mechanisms.
In addition to promoting SidM/DrrA for Rab1 recruitment, LidA may have other function(s). LidA interacts with the active conformation of Rab1 and exhibits no enzymatic activity [39]. Structural comparison showed that binding of LidA to Rab1 shares some features of the interaction between the effector protein Rabenosyn-5 with Rab4 [54]. Furthermore, recruitment of Rab1 is essential for delivering ER-derived vesicles to LCV [42,43,62]. Collectively, these results suggest that one potential function of LidA is to act as an effector of Rab1 and signal downstream components for remodeling of LCV. LidA was previously predicted and further confirmed in the present study to be a coiled-coil rich protein, one type of the best characterized Rab tethering factors [27,54,63]. The terminal parallel helix forms an antiparallel long coiled-coil and extends to the outside of interacting region ( Figure 2C). Secondary structure analysis suggests that in full-length LidA, the length of this domain is likely to be longer than 100 Å . It is thus possible that LidA may function as a tethering factor, bridging the ER-derived vesicles and the LCV [49]. One potential advantage of employing its own protein as a tethering factor would allow L. pneumophila to be selective for recruiting ER-derived components for its growth. It is well established that a GTP-bound form of Rab GTPases interacts with their effectors for signaling. The only exception to this rule is Protrudin that interacts with the GDP-bound form of Rab11, regulating Rab11-dependent membrane recycling to promote the directional membrane trafficking [64]. While interacting with GDP-Rab1, LidA appears to be different from protrudin in that it essentially recognizes the active conformation of Rab1 and probably other Rabs [39,43].
During the peer review of this manuscript, the structure of Rab8-LidA complex was published online. Structure comparison shows that Rab8-LidA complex adopts quite similar conformations with our structure and that the binding interface is highly conserved, consistent with our proposal that lidA recognizes Rabs in a similar way. We note that the authors report a binding affinity of LidA for GTP-bound Rab1 that is roughly one order of magnitude stronger than what we measured using ITC. At this point, it is inconsistent with our ITC data that LidA binds to both GDP-and GTP-Rab1 with a nearly equal affinity.

Protein expression purification
The genes of lidA and sidM were amplified from genomic DNA of Legionella pneumophila strain Corby. The DNA fragment encoding human Rab1a (noted Rab1 throughout the text) and other 14 kinds of Rab family members were amplified from a homemade human cDNA library. All the genes encoding the mutant proteins were produced using a two-step PCR procedure. LidA, both full-length (FL) and various fragments, SidM variants include 317-545 and 1-545 were subcloned into the pET15b (Invitrogen), Rab1 variants and Rab2, 4,6,7,8,11,14,18,23,27 were subcloned into the pGEX-6P-1 plasmid (GE Healthcare), Rab5, 9, 20, 22 were subcloned into both of the two vectors. Each protein was produced in E. coli BL21 (DE3). Cells grow at 37uC until the OD 600 reach to 0.8. LidA proteins were induced with 0.5 mM isopropyl-b-D-Thiogalactopyranoside (IPTG) for 15 h at 15uC, Rab1 and other Rabs were all induced with 0.2 mM IPTG for 15 h at 22uC. All proteins were purified using affinity, anion exchange and gel filtration chromatography. To obtain LidA-Rab1 complexes, LidA and Rab1 proteins were mixed up with a 1:1 molar ratio at 4uC for 2 hours. The LidA(188-449)-Rab1(1-191) complex and LidA(224-559)-Rab1(S25N; 1-176) complex were further purified using the above mentioned procedure except the GST tag and His tag were removed by homemade PreScission protease digestion before the anion exchange step. Both of the protein complexes have a final concentration of 10 mg/mL. All purification processes were performed at 4uC unless noted otherwise.   complex diffracted to 2.2 Å . Data sets were processed using the HKL2000 software suite [65].

Structure determination
The crystal structures of the two kinds of LidA-Rab1 complexes were determined by molecular replacement using PHASER [66] with the coordinates of Rab1 as the searching models (PDB ID code 2FOL) [67]. The atomic models were built using COOT [68] and refined using PHENIX [69]. Data collection and structure refinement statistics are summarized in Table S1. A SA-composite omit map of LidA index finger shows the good match between the atom skeleton and the electron density map as supplementary Figure S6. All the molecular graphics Figures were generated using PyMol (http://www.pymol.org).
In vitro pull down assay 15 kinds of purified Rab family members (WT; FL) with Nterminal GST tag (His-fused Rab20 is exceptional) was immobilized onto 200 mL glutathione-Sepharose resin (GE) or Nichelating sepharose (GE healthcare) for His-Rab20. The resin was then washed three times to remove excess unbound protein.
Untagged LidA(FL) protein solution was loaded onto the Rabsimmobilized beads and incubated at 4uC for 2 h. The loaded protein dose of LidA and Rabs was at a molar ratio of 2:1. The resin was washed three times using the wash buffer (25 mM Tris-HCl buffer pH 8.0, 100 mM NaCl) to remove unbound LidA. Proteins binding to the resin were eluted by elution buffer (25 mM Tris-HCl buffer pH 8.0, 100 mM NaCl, 3 mM DTT, 5 mM GSH for glutathione-Sepharose resin and 25 mM Tris-HCl buffer pH 8.0, 100 mM NaCl, 250 mM imidazole for Ni-chelating sepharose). All samples were subjected to SDS-PAGE which was visualized by Coomassie Brilliant Blue staining.

Accession numbers
Crystal structure of LidA(224-559)-Rab1(S25N; 1-176) complex and the structure factor have been deposited in the Protein Data Bank (http://www.rcsb.org/pdb) under ID codes 3SFV. Crystal structure of LidA(188-449)-Rab1(Q70L; 1-191) complex and the structure factor are deposited with access codes 3TKL. Protocol S1 Domain mapping by gel filtration. To map the interaction domain of LidA with Rab1, the proteins of LidA(FL) and fragments, Rab1(FL, Q70L) were purified by affinity chromatography with N-terminal His-tag, then mixed protein solutions together at 4uC for 2 h, the protein concentration of Rab1 and LidA was at a molar ratio of 1:1. Afterwards, the protein mixtures were subject to size exclusion chromatography on a Superdex-200, monitored by UV absorption at 280 nm. The buffer of gel filtration containing 25 mM Tris-HCl (pH 8.0), 100 mM NaCl, and 3 mM DTT. Each aliquots of the peak fraction were subjected to SDS-PAGE which were visualized by Coomassie Brilliant Blue staining.