Mechanism of effector capture and delivery by the type IV secretion system from Legionella pneumophila

Legionella pneumophila is a bacterial pathogen that utilises a Type IV secretion (T4S) system to inject effector proteins into human macrophages. Essential to the recruitment and delivery of effectors to the T4S machinery is the membrane-embedded T4 coupling complex (T4CC). Here, we purify an intact T4CC from the Legionella membrane. It contains the DotL ATPase, the DotM and DotN proteins, the chaperone module IcmSW, and two previously uncharacterised proteins, DotY and DotZ. The atomic resolution structure reveals a DotLMNYZ hetero-pentameric core from which the flexible IcmSW module protrudes. Six of these hetero-pentameric complexes may assemble into a 1.6-MDa hexameric nanomachine, forming an inner membrane channel for effectors to pass through. Analysis of multiple cryo EM maps, further modelling and mutagenesis provide working models for the mechanism for binding and delivery of two essential classes of Legionella effectors, depending on IcmSW or DotM, respectively.

L egionella pneumophila is an opportunistic human pathogen that causes a type of severe pneumonia called Legionnaire's disease 1 . It also has natural hosts among protozoa 2 . The bacterium translocates into the cytosol of the host a plethora of effector proteins that hijack cell functions to create a specialised organelle that supports intracellular replication 3 . L. pneumophila effectors are injected into the host using a T4S system 4,5 , which is encoded by 27 genes of the dot/icm gene cluster including 3 ATPases namely DotO, DotB, and DotL 6 . In the study presented here, we focus on L. pneumophila DotL, a membrane-embedded AAA + T4S system ATPase and member of the VirD4 family of proteins 7 . In Legionella, DotL is part of a large complex that includes the proteins DotM and DotN ( Supplementary Fig. 1a). Together, they form the Type IV coupling complex (T4CC) 8 . The T4CC contains several types of binding sites that recruit different classes of effectors. Indeed, depending on their mode of recruitment, effectors can be grouped into two classes: (a) effectors that are dependent on their binding to a complex of the proteins IcmS and IcmW (IcmSW) 9 , and (b) the other that include effectors which are IcmSW-independent. Among the latter, there is a subset of effectors that contain a C-terminal secretion signal sequence rich in Glu residues (referred to as "Glu-rich signal peptide or Glu-rich SP") 10 . Effector-bound IcmSW binds to T4CC through binding to the very C-terminal sequence of DotL ( Supplementary Fig. 1a) [11][12][13] . In contrast, the subset of acidic Glu-rich SP-containing effectors bind to DotM 14 . Although some single component fragments of the T4CC are structurally characterised 12,14 , there is no overall view of the intact, fully assembled complex. Here, we present the atomic structure of an intact, fully-assembled Legionella T4CC.

Results and discussion
Purification and composition of the T4CC. We purified the T4CC from Legionella cell membranes after solubilisation with detergents and taking advantage of a Strep-tag inserted at the Cterminus of DotL (Methods and Supplementary Tables 1, 2 and  3). The complex not only contains DotL, DotM and DotN but also 5 additional proteins (Fig. 1a): IcmS, IcmW, LvgA, and two previously-uncharacterised proteins encoded by two annotated open reading frames, lpg0294 and lpg1549. We named these proteins DotY and DotZ, respectively, since they co-purify with the dot/icm Legionella T4CC. The size of this complex is 300 kDa as assessed by SEC-MALS, consistent with a complex that may contain 1 copy each of the 8 proteins. The presence of DotY and DotZ (the encoding genes of which are located outside the Legionella dot/icm gene cluster) was unexpected. To assess their role, three deletion mutants were made, ΔdotY, ΔdotZ and ΔdotYZ, where the dotY or dotZ or both genes were deleted, respectively. Intracellular growth of these mutants in the protozoa Acanthamoeba castellanii was reduced (47(±11), 36(±26), and 25 (±15)%, respectively, compared to wild-type), consistent with a previous report of a transposon-insertion lpg0294 (dotY) mutant 15 . Complementation of the ΔdotY or ΔdotZ strains with wild-type dotY or dotZ gene, respectively, restored growth to wild-type levels ( Supplementary Fig. 2), indicating that reduction in intracellular growth in the mutants is due to deletion of the targeted gene(s). These data are similar to results obtained for the mutants that are deficient in the coupling protein chaperone proteins IcmS, IcmW, and LvgA 16,17 . These mutants display relatively minor defects in effector translocation assays, that result in more pronounced defects in intracellular replication. Since the intracellular replication defects are the result of decreased effector translocation, complementation studies typically use intracellular replication assays to confirm the absence of a secondary mutation, which confirms that effector translocation defects have been restored.
Next, effectors translocation was monitored using Cya-fusions of 5 different effectors: RalF, a well-characterized effector, Lem21 and LegC8, which are known IcmSW-dependent effectors 14 , Lpg1663 and CegC3 which are acidic Glu-rich SP-containing effectors that we have recently shown to be recruited by DotM 14 . We show here that translocation of these effectors into CHO cells was affected significantly by dotY or dotZ deletions (5-to 10-fold reduction compared to wild-type (Fig. 1b)). The Intracellular growth complementation results confirm the decreased levels of effector translocation are due to the genes deletions and not due to secondary mutations.
Finally, we show that DotY and DotZ are not themselves translocated (Fig. 1c). We conclude that DotY and DotZ are integral parts of the T4CC and play significant roles in the translocation of most effectors that we have tested. They are also unique to the Legionella genus.
Structure of the DotLMNYZ hetero-pentameric complex. The structure of the T4CC was next determined using cryo-electron microscopy (cryo-EM). Single particle reconstruction yielded a map with an average resolution of 3.7 Å but 3.5 Å in many parts (Supplementary Fig. 1b-f). The quality of the electron density was sufficient ( Fig. 1d) to solve the structures of DotL and DotM except for their trans-membrane segments, DotN, DotZ, and the first N-terminal 77 residues of DotY (Figs. 1e and 2, and Supplementary Table 4). These parts constitute the "hetero-pentameric core" of the T4CC. This core structure has the shape of a right angle triangle with a longer and short side of 14.2 and 9.1 nm, respectively (Fig. 1e). The transmembrane segments of DotL and DotM locate at the end of the short side (Fig. 1e). The structure of the T4CC hetero-pentameric core reveals a large interaction network involving 18,146 Å 2 of buried solvent accessible surface area, bringing all 5 proteins together through 8 interfaces (Figs. 2  DotL plays a central role in the assembly of the T4CC, forming the 1 st and 2 d largest interfaces, with DotM (3089 Å 2 in each protein) and DotN (1756 Å 2 ), respectively. DotL has a long Cterminal tail which starts at residue 572 in the region of the structure proximal to the membrane just after the β11 strand and runs down the entire T4CC structure (Figs. 1e, 2 and 3a-d, and Supplementary Figs. 3a and 4b-d). It encompasses a long segment termed "β11α14" (Fig. 3b, c), which is so-called because it lies between β11 and α14 (thereafter, all regions between secondary structures will be referred to in a similar way). This region of the tail is an integral part of the DotL structure and makes extensive interactions with DotM, notably running within a groove formed by two sub-domains of DotM (Fig. 3c). It is followed by helix α14, a loop between α14 and α15, and finally α15 and α16: this region forms the interface with DotN 12 (Fig. 3d and Supplementary Figs. 3c and 4c). Other regions of DotL involved in DotM binding include residues in the region of the β1 and β2 strands, residues in the α2β4, α3β5 and β5α4 regions, as well as in the region at and around α10 (α9α10 and α10β6) and in α11β7. On the DotM side, residues involve in contact with DotL are situated in the α1 to α3 region at the N-terminus of the protein, and in the C-terminal half of the protein from η1 to α14 (Fig. 3b, Supplementary Figs. 3a, b and 4b). Likewise, the interactions stabilising the recruitment of DotN to the T4CC core extend beyond the DotL C-terminal tail. DotN is indeed further involved in contact with DotM, together forming with this protein the 5th largest interface between T4CC core proteins (659 Å 2 in each protein). Residues in α4 and α12α13 regions of DotM interact with residues in α1, α1α2, α3β3, and α6 of DotN DotZ lies along the long side of the T4CC's right triangle shape. This 294 amino acid protein has an elongated structure, the core of which is made of three long α-helices (α3, α4, and α7) (Figs. 2 and 3f). Four smaller helices, α1, α2, α5, α6 wrap around the distal end of the α3-7 bundle, α1 departing at a 90°angle to form together with α3 a groove (Fig. 3f). It makes significant contact with DotN (1517 Å 2 , the 3rd largest interface) and DotY (1396 Å 2 , the 4th largest), but engages weakly with DotM (355 Å 2 ) and DotL (177 Å 2 ) (Supplementary Figs. 3a-e and 4d, f-h). DotN interactions with DotZ include residues of the very Cterminal helix (α8) and residues in α4 and the α4α5 loop of DotN which inserts into the DotZ groove (Fig. 3g). DotM interacts with DotZ through an extended loop, β1β2. β1β2 of DotM, α7 and α8 of DotZ, and α3β3 of DotN form a cavity (Figs. 2 and 3h and Supplementary Fig. 4h).
For DotY, clear density is only observed for three N-terminal helices (α1-3) and a substantial loop following α3 (Figs. 2 and 3i). These regions of DotY are ordered because they make extensive interactions with DotZ. This interface brings together residues in α1 and α2 of DotY with residues in α3 and α4 of DotZ and a cluster of residues in the DotY α3 loop with the DotZ groove ( Fig. 3i and Supplementary Figs. 3d,e and 4g). Proximity of DotN and DotY within the DotZ groove leads to interactions between these two proteins ( Supplementary Fig. 4i).  45 . Bar shape-coding for each mutant and wild-type strains is indicated on the right. For the ΔdotY, ΔdotZ, ΔdotYZ mutants, differences in translocation levels were found to be significant with a P value of <0.005 comparing to WT. Only for RalF, P values were <0.05. c Translocation of DotY and DotZ. Translocation of Cya-DotY and Cya-DotZ was compared to Cya-RalF, and also assayed in the ΔT4SS strain. Bar color-coding is shown at the top. For both Cya-DotY and Cya-DotZ, there were no significant differences in translocation levels between the WT and ΔT4SS strains. d Electron density of the T4CC hetero-pentameric core. The map was contoured at 7 σ level. Color-coding is per protein, red, cyan, blue, orange yellow and green for DotL, DotM, DotN, DotZ and DotY, respectively. e Two views of the structure of the T4CC hetero-pentameric core. The two views are related by 180°rotation. At left, DotL, DotM, DotN, DotZ and DotY are shown in ribbon, color-coded red, cyan, blue, orange yellow, and green, respectively. Electron density is as in panel d, but semi-transparent. Complex dimensions are indicated. At right, the proteins are represented in ribbon except for the C-terminal tail of DotL, which is shown in surface representation. Color-coding is as shown in the panel at left. The locations of the disordered transmembrane (TM) regions of DotM and DotL are indicated by a cyan and red cylinders, respectively, thereby providing the location of the inner membrane (IM). For (b) and (c), data are representatives of three independent experiments (n = 3), each strain with biological triplicates. For each independent experiment, effector translocation values in mutants were normalized against their ratio to WT. Graphs report mean intracellular cAMP levels ± standard deviation for each strain. Indicated P values are mutant strains in comparison to wild-type, calculated by two-tailed Student's t test. NS not significant (P > 0.05). For (a-c), Source data are provided as a Source Data file. Validation of the T4CC hetero-pentameric structure. In order to validate the structure of the T4CC hetero-pentameric core, six regions involved in interfaces between two proteins within the complex structure were targeted for mutations (named M1-M6). These regions were chosen because of the multiplicity of contacts that they make. The location and structure of these regions are shown in Fig. 4a. M1 comprised the mutations T205R, L208R, Y211R in DotM α4α5 loop at the interface of DotM and DotN. For M2, the C-terminal region from residues 200 to the Cterminus of DotN corresponding to the C-terminal half of α8, which makes contact within the wedge of DotZ, were deleted. Another deletion mutant, M3, removed DotZ residues from residue 283 (just after α8) to the C-terminus. This region interacts with α8 of DotN and β1β2 of DotM. Two mutants, M4 (Q326R, T327R) and M5 (A363R, E364R, D366R), in DotM and DotL, respectively, were obtained, each affecting the contacts between residues in α11α12 in DotM and residues in α9α10 of DotL. Finally, another mutant (M6) was obtained by mutating another contact region between DotM and DotL at residues V300R, P302R, and S303R of the DotM η2η3 loop. All mutations were introduced as described in Methods. All mutants were tested for their ability to grow intracellularly in A. castellanii as described above for the dotY and dotZ deletion mutants (Table 1) (Table 1). We conclude that the interfaces observed in the T4CC core structure provide an accurate account of the interactions taking place in vivo within the complex.
The T4CC as a hexamer of hetero-pentameric units. DotL belongs to the VirD4-family of AAA + ATPases which may purify as monomers but all function as hexamers 18 . Moreover, the only known structure of a representative VirD4 protein is that of TrwB, encoded by the R388 T4S system, and this structure is hexameric 19 . Thus, we used the TrwB structure to model a hexamer of DotL and consequently a hexamer of the Legionella T4CC hetero-pentameric core (Fig. 5). The resulting structure resembles a starfish, 26 nm in diameter (Fig. 5a). DotL forms a channel with an inverted funnel shape, constricted (2.0 nm) at the base and flaring up to 4.3 nm near the membrane (Fig. 5a). Preceding the DotL channel lays a 6.7 nm diameter chamber, wide open on the cytosolic side but only accessible through the 2 nm DotL constriction on the membrane side (Fig. 5a). The interface between adjacent DotL molecules is extensive with circa 2,000 Å 2 of buried surface area in each DotL subunit. It is essentially similar to that of TrwB and therefore will not be described here (see however Supplementary Figs. 3f and 4j, k for details). A published mutational study of DotL 11 identified 12 residues across the DotL sequence, which, when mutated, resulted in intracellular growth defects. Three of these residues locate at the proposed DotL-DotL interface in the T4CC hexamer (Fig. 4b). These mutants provide validation for the proposed interface involved in hexamer formation, indicating its functional relevance in vivo. In the T4CC hexameric form, the membrane regions of both DotL and DotM would be expected to form a trans-membrane channel. Hexameric assembly might be induced by effector binding or a constitutive hexamer might be formed in vivo in the membrane environment. Kwak et al. 12 also proposed a hexameric model based on TrwB. Our model differs considerably. That's because the hexameric model proposed by Kwak et al. is based on multiple structures of separate subparts and some elements (DotM) were missing. In contrast, our model is based on the structure of an intact, fully assembled, complex. Differences include the following: the structure of DotL is here solved; DotM's position is experimentally established within the T4CC; IcmSW does not locate next to DotN but instead protrudes out; and we show that the T4CC actually include two additional proteins, DotY and DotZ.
The IcmSW positional flexibility. Single particle analysis of the T4CC revealed a U-shape density reminiscent of the structure of IcmSW bound to a DotL C-terminal peptide 12 (termed "IcmSW-DotL 672-783 "; Supplementary Fig. 5a, b). Focusing on this U-shape density, we obtained a 9.7 Å resolution reconstruction into which the crystal structure of IcmSW-DotL 672-783 could easily be docked (correlation coefficient of 0.92; Supplementary Fig. 6). We concluded that the U-shape density that we observed does indeed correspond to IcmSW bound to DotL 672-783 . LvgA, which copurifies with the T4CC and is known to bind IcmSW 12 , is absent from the structure determined here, likely because it is either too flexible or dissociates upon freezing during grid preparation, a well-known effect on protein complexes 20 .
We next aimed to generate reconstructions of the entire T4CC including IcmSW. Three-dimensional (3D) classifications allowed us to resolve multiple orientations of the IcmSW module relative to the hetero-pentameric core (Fig. 5b, c and Supplementary  Fig. 5c). IcmSW's positional flexibility is afforded by the DotL residues immediately preceding the IcmSW-binding region. In the context of the hexamer model, maps superposition shows that the IcmSW module swings at the base of the structure in a trajectory that directs the module in and out of the DotL channel (Fig. 5b, c).
Acidic Glu-rich SP effector binding. As mentioned before, one subset of effectors does not rely on IcmSW for transport, but instead binds directly to DotM 14 . These effectors are characterised by a particularly acidic Glu-rich SP. Previously, we characterised the surface of DotM involved in binding of Glu-rich SP and generated a structural model of the DotM-SP interaction 14 . When superposing this model onto the T4CC heteropentameric core structure using the DotM structures in both (Fig. 5d), the N-terminal end of the Glu-rich SP is observed inserting within the cavity mentioned above formed between DotM, DotZ, and DotN (Figs. 2, 3h and 5d). In the context of the T4CC hexamer (Fig. 5d), it can be seen that, by going through the cavity, the peptide reaches out to the DotL channel.
Models for IcmSW-and DotM-dependent transport. The results presented here provide mechanistic models for recruitment and delivery of two types of effectors, the IcmSWdependent class and the acidic Glu-rich SP one. These models are detailed in Fig. 6a, b. The positional flexibility that we observe for IcmSW would provide a means for this module to scan the environment and maximise effectors capture while the defined trajectory we observe will facilitate their delivery to the DotL channel. On the other hand, Glu-rich SP-containing effectors bind to a region of DotM that induces the SP to insert into a cavity formed by three of the T4CC proteins. For the full-length The interface between two adjacent DotL subunits in the proposed T4CC hexamer (shown in red and orange ribbons). Three mutations reported by Sutherland et al. 11 to affect intracellular growth locate to this interface. The Cα atom of these residues is shown as a sphere. The residues as well as the secondary structures they belong to are labelled.  Fig. 4a) of the complex interfaces as described in the main text. Mutants were tested for their ability to grow intracellularly in Acanthamoeba castellanii (AC) as described in Methods. Data are presented as mean values + /− standard deviation. Source data are provided as a Source Data file.  (2) it is weakly anchored to DotZ (Fig. 3g). Once the DotM-bound effector has slotted into place through the open cavity, it would lie under the DotL channel, ideally positioned for translocation. Further steps beyond binding and delivery may include partial unfolding 21 by either IcmSW for IcmSW-dependent effectors in a way reminiscent of type III secretion chaperones 22,23 or by DotL itself for other effectors. The structure of the T4CC from L. pneumophila reveals a remarkably versatile, multi-site, recruitment and delivery platform. Given the large number of effectors Legionella is able to secrete 3,24 , it is likely that other capturing and delivery mechanisms involving the T4CC will be unravelled in the future. Such a large structure may indeed contain additional sites for binding of other types of effectors which are neither IcmSW dependent nor dependent on DotM. Interestingly, the T4S system is not the only secretion systems endowed with a large platform for effector recruitment and delivery 25 . Thus, our structural investigation might provide a potential paradigm on effector recruitment by other secretion machines. Finally, it could be argued that such a multi-site effector-binding platform might provide scope for temporal regulation of effector secretion, some binding sites functioning in early stages of secretion while others coming in later.

Methods
Bacterial strains and constructs. Strains, plasmids and oligonucleotides used in this study are shown in Supplementary Tables 1, 2 and 3, respectively.
Isogenic Lp02 strains were produced as previously described 26 . To generate the DotL-Strep tag construct, dotL was first cloned into suicide plasmid pSR47S with 1000 bp upstream and downstream of the gene's 5′ and 3′ sequences. Then, the sequence encoding a Strep-tag (SASWSHPQFEK) was introduced to the 3′-end of dotL.
For production of the KO strains Δlpg0294 (DotY) and Δlpg1549 (DotZ) in the Lp01 background, genes were cloned with 1000 bp upstream and downstream to pSR47S, later deleted leaving a double stop codon TAA after the first ATG, followed by 20-25 bp of the gene's 5′ and 3′ sequences. For production of the double knockout strain, after creation of Δlpg0294 strain, additional mutagenesis was performed with the Δlpg1549 construct. All strains were verified by colony PCR.
For DotY/DotZ complementation assays, dotY and dotZ were cloned into pJB1806 plasmid using InFusion. Wild-type dotY and dotZ were cloned into the pJB1806 backbone with 200 bp upstream and downstream, so that their native promotor is used for expression.
Interface mutations were introduced to the pSR47S constructs of dotL, dotM, DotN, and dotZ by In-Fusion or by Quick change. First attempt of mutagenesis was conducted in Lp01 wild-type background, and in cases of no positive hits (0/70 colonies), strains were generated in the background of Lp01 ΔdotB, a strain with inactive T4BSS. All mutated strains were verified by colony-PCR, followed by sequencing of the mutated region.
For DotB complementation assays, dotB fused at its 3′-end to the Strep-tag encoding sequence mentioned above was cloned into pMMB207 plasmid using InFusion.
For DotL Strep purification, 48 h heavy patch cells were inoculated and grown for additional 26 h in AYE medium and supplements to achieve a final OD 600 of 3.2-3.6. Cells were harvested and resuspended in buffer LPA (50 mM Tris pH 8.0, 0.2 M NaCl, 2 mM EDTA, 20 mM MgSO4) and 0.5 M sucrose, 0.1 mg/ml lysozyme, DNAse I and protease inhibitor (PI) (Roche). After rotation for 45 min at 4°C, cells were spun down and then re-suspended in buffer LPB (50 mM Tris pH 8.0, 2 mM EDTA, 20 mM MgSO4, and PI), followed by 3 rounds of high pressure (40,000 psi) homogenisation. The lysate was centrifuged at 17,300 × g for 20 min to remove the cell debris, followed by ultracentrifugation at 167,000 × g for 2 h. For each 1 gr of membranes, 6 ml of LPA buffer was added and membranes were frozen at −80°C until needed. An additional 6 ml of re-suspension buffer LPA and 2.5% DDM (n-Dodecyl-β-D-Maltopyranoside, Anatrace) was added after thawing. Solubilized membranes were gently shaken at room temperature for 2 h, then centrifuged again at 142,000 g for 30 min to remove insoluble materials.
Soluble membranes were loaded on 5 ml StrepTrap column (GE Healthcare), followed by extensive wash in LPA buffer and 0.05 % DDM, and eluted in LPA buffer with 0.05% DDM and 2.5 mM desthiobiotin (Sigma). Eluted fractions were analysed by SDS-PAGE, pooled, concentrated and loaded on SEC column Superose 6 (GE Healthcare). Peak fractions were pooled, concentrated, and protein concentration was determined by OD 280 measurement. To remove DDM, the  Fig. 6 Mechanistic models of substrate recruitment and delivery by the Legionella T4CC. a IcmSW-dependent effectors. Positional flexibility of the IcmSW module relative to the rest of the structure provides a means to scan the immediate environment to maximise substrate capture (step 1) by the T4CC. Once bound, because of the motions trajectory, the effector will be delivered to the DotL channel (step 2 "delivery"). Binding to IcmSW may induce IcmSW-dependent effectors to partially unfold 23 . However, structural details of effector-IcmSW interactions are not known. Additionally, DotL might be able to use its ATPase activity for the dual purpose of unfolding and transport (step 3 "translocation"). LvgA is not shown because it is not present in the structure presented here. b Glu-rich SP-containing effectors. Glu-rich SP-containing effectors bind to a region of DotM that induces the SP to insert into a cavity formed by three of the T4CC proteins. For the full-length effector protein to bind there, there would be a requirement for the cavity to open up (step 1 "capture"). To do so, we hypothesize that the β1β2 loop of DotM might swing out, allowing the DotM-bound effector to slot in to position itself under the DotL channel (step 2 "delivery"). How would then the effector in the DotL channel unfold remains unclear but it cannot be excluded that DotL might use its ATP-driven power to mediate unfolding (step 3 "translocation").
concentrated complex solution was incubated with Amphipol A8-35 (Anatrace) at 1:5 ratio for 4 h, followed by overnight incubation with biobeads (Biorad). The sample was then reloaded on the Superose 6 column, and peak fractions were collected and concentrated for cryo-EM studies.
Cryo-EM grid preparation and data acquisition. Aliquots of the purified T4CC were applied to negatively glow discharged UltrAuFoil R1.2/1.3 grids (Quantifoil, Germany) and vitrified in liquid ethane using a Vitrobot Mark IV (Thermo Fisher, USA) at 4°C and 94% humidity. The data were collected at the eBIC National facility (Diamond Light Source, UK) and ISMB Birkbeck EM facility using Titan Krios microscopes (Thermo Fisher, USA) operated at 300 keV and equipped with a Quantum energy filter. The images were collected with a post-GIF K2 Summit direct electron detector operating in counting mode, at a nominal magnification of 130,000, corresponding to a pixel size of 1.047 Å. An energy slit with a width of 20 eV was used during data collection. The dose rate on the specimen was set to 4.9 e per pixel per second, and a total dose of 54 e Å −2 was fractionated over 48 frames. Data were collected using EPU software (Thermo Fisher, USA) with a nominal defocus range set from −1.5 μm to −3.5 μm. A total of 19,491 micrographs were collected.
Cryo-EM data processing. RELION 3.0 29 was used for motion correction and dose weighting with MOTIONCOR2 30 followed by CTF estimation using CTFFIND v4.1 31 . An initial low-resolution map was obtained using RELION 3.0 following the workflow described in Zivanov et al. 32 . Reprojections of this map were used to pick particles with GAUTOMATCH v0.56 33 . Dataset was subjected to multiple rounds of 2D and Ab-initio classifications with CRYOSPARC v0.6.5 34 leading to selection of 626,230 out 8,702,486 particles. Selected particles were re-extracted from 16,861 micrographs using RELION 3.0, following by 3D refinement and 3D classification that resulted in further selection of 541,522 particles. These particles were re-centered, used for 3D refinement with a mask focusing on the DotLMNYZ density, and subjected to 3D classification with the same mask without image alignment using Tau = 20. The two best resulting classes corresponding to 241,838 particles were selected. To limit anisotropy and improve the quality of the map,~20,000 particles corresponding to preferential views were removed from the star files using rlnMaxValueProbDistribution criteria. The final subset of 219,593 particles was imported to CRYOSPARC v2.9.0, to perform Non-Uniform Refinement that resulted in an electron density map with a nominal resolution of 3.7 Åas estimated using gold standard Fourier shell correlation (FSC) with a 0.143 threshold (Supplementary Fig. 1 and Supplementary Table 4). This map was AutoSharpen using PHENIX v1.14 35 .
To determine the structure of the U-shaped domain and ascertain that it corresponds to IcmSW, we selected particles with the characteristic U-shape, either alone or attached to the T4CC core. Indeed, a small fraction of the U-shape density was observed detached from the T4CC core. The 541,522 particles set selected during the T4CC core structure determination (see above) were re-extracted using RELION 3.0 with a shift and centering on the U-shape feature. 3D classification was performed using parameter Tau = 2 without alignment, and 103,532 particles corresponding to 4 classes showing U-shape density were selected. In addition, a set of 64,750 particles with the characteristic U-shape were selected after 2D classification and ab-initio 3D classification using CRYOSPARC. These two sets of particles were then combined and subjected to 2D classification and ab-initio classification using CRYOSPARC v0.6.5. A final subset of 18,210 particles was selected and subjected to homogeneous refinement. The resulting 9.7 Åresolution map (as estimated by the gold standard FSC with a 0.143 threshold ( Supplementary  Fig. 6)) was sharpened using CRYOSPARC v0.6.5 with a B factor value of −1500. The DotL-Cter IcmSW crystal structure (PDB ID 5×1E) was docked as a rigid body into the final map using CHIMERA v1.13.1 36 .
Different positions of IcmSW relative to the hetero-pentameric T4CC core were resolved using CRYOSPARC ab-initio classification with the high-resolution limited to 20 Å( Supplementary Fig. 5). Initially 10 classes were obtained using the subset of 541,522 particles selected during the T4CC core structure determination. Classes showing the IcmSW domain in the same position relative to the T4CC core were selected and combined. Ab-initio classification and selection of classes were repeated two more times. All maps showing a clear density for the IcmSW domain were aligned using the hetero-pentameric T4CC core region and CHIMERA v1.13.1. Seven maps were selected to represent the extent of IcmSW domain motion relative to the hetero-pentameric T4CC core (Fig. 5b, c).
Model building and refinement. I-TASSER 37 was used to generate a model of the DotL ATPase domain (100-589) derived from the TrwB structure (PDB entrey 1GKI). This model was combined with the DotL part of the DotL 590-659 -DotN crystal structure solved previously (PDB ID 5 × 42 12 ) to generate the starting model of DotL 104-658 . For DotM and DotN, the previous crystals structures by Meir et al. 14 and Kwak et al. 12 were used (PDB IDs 6EXD and 5×42, respectively). DotY and DotZ were built de novo in COOT v0.8.9.1 38 based on the density map and secondary structure prediction (PSIPRED 4.0 39 ). Simulated annealing in the initial rounds of real-space refinement with PHENIX was used.
The entire structure of the T4CC hetero-pentameric core (DotLMNYZ) was improved by iterative rounds of manual adjustment in COOT v0.8.9.1 followed by real-space refinement in PHENIX v1.14 using secondary structure restraints. MOLPROBITY v4.4 40 was used to evaluate the quality of the structures. All data and model statistics are reported in Supplementary Table 4.
Interaction analysis was conducted using PISA server 41 , and structure representations were generated using UCSF CHIMERA v1. 13 Cell Culture. CHO FcγRII cells 43 used for translocation assays were cultured at 37°C in 5% CO 2 in RPMI-1640 plus 10% FBS. A. castellanii (ATCC 30234) were cultured routinely at room temperature in ATCC medium 712 (PYG).
Legionella intracellular growth in eukaryotic hosts. Intracellular growth assays were performed as previously described 44 . Specifically, A. castellanii were infected in AC medium. Cells were plated at 2 × 10 5 cells/ well and incubated at 37°C 2 h prior infection. Two-day heavy patch bacterial strains were grown on CYE plates with appropriate antibiotics (100 μg ml −1 streptomycin for WT and mutant strains, supplemented with 10 μg ml −1 chloramphenicol for the strains containing the complementing plasmids). Bacterial strains were added to A. castellanii plates at MOI of 0.1 (2 × 10 4 cells per well) followed by centrifugation for 5 min at 350 × g at room temperature and incubation at 37°C for 1 h. For intracellular growth of the interface mutants, A. castellanii were infected in AC medium containing 1 mM IPTG. A ΔdotB strain complemented by pMMB207:DotB with or without IPTG in the media was included as positive and negative control, respectively.
CYA assay. Cya assays were conducted as previously described 14 . Specifically, CHO FcγRII cells (1 × 10 5 cells per well) were placed into 24-well tissue culture plates in α-MEM plus 10% FBS 1 day prior to infection. On the day of infection, 2days heavy patch of Legionella strains (W.T., ΔT4SS, ΔdotY, ΔdotZ, or dotYdotZ double knockout mutants) transfected with the Cya-containing plasmids were diluted into α-MEM plus 10% FBS medium supplemented with Rabbit anti-Legionella antiserum diluted at a ratio of 1:1000 (which facilitates Legionella adhesion) and 0.5 mM IPTG (to induce Cya fusions), and incubated at R.T for half an hour prior to infection. The CHO FcγRII cell culture medium was aspirated before adding to each well the corresponding Legionella strains (3.0 × 10 6 bacteria per well). The plates were centrifuged onto a confluent monolayer of host cells for 5 min at 200 × g. Plates were immediately warmed in a 37°C water bath for 5 min, then placed in a CO 2 incubator for a total of 1 h. Cells were washed three times with ice-cold PBS and lysed in 200 μl of extraction solution (50 mN HCl/0.1% Triton X-100) on ice. After boiling for 5 min, extracts were neutralized with 12 μl of 0.5 M NaOH and cAMP was extracted with 2 volumes of ethanol. Insoluble materials were pelleted by centrifugation, and the soluble materials containing cAMP were lyophilized. The cAMP levels were determined for each extract by using an ELISA kit according to manufacturer's instructions (Amersham Biosciences, RPN-225).
Statistical analysis. Statistical analysis was performed with GraphPad Prism v.5.0 (GraphPad Software, La Jolla, CA, USA). For comparison of two groups, Lp01 WT against mutants, an unpaired t test was employed. A P value of <0.05 was considered statistically significant. All experiments were performed at least three times, each strain with three biological triplicates. The data are expressed as mean ± standard deviation (s.d.). For the Cya translocation assays, effectors translocation values in mutants were normalized against their ratio to WT. For the Cya effector translocation assay, we also employed one sample t test Vs. 1 to determine significance between WT and mutant strains. P values were calculated and still showed a significant difference.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
Entry codes for the EM density map and the atomic model of the hetero-pentameric T4CC core are EMD-10350 and PDB ID 6SZ9, respectively [https://doi.org/10.2210/ pdb6SZ9/pdb]. The source data underlying Fig. 1a-c, Table 1, and Supplementary Fig. 2 are provided as a Source Data file. Source data are provided with this paper.