The Polar Legionella Icm/Dot T4SS Establishes Distinct Contact Sites with the Pathogen Vacuole Membrane

ABSTRACT Legionella pneumophila, the causative agent of Legionnaires’ disease, is a facultative intracellular pathogen that survives inside phagocytic host cells by establishing a protected replication niche, termed the “Legionella-containing vacuole” (LCV). To form an LCV and subvert pivotal host pathways, L. pneumophila employs a type IV secretion system (T4SS), which translocates more than 300 different effector proteins into the host cell. The L. pneumophila T4SS complex has been shown to span the bacterial cell envelope at the bacterial poles. However, the interactions between the T4SS and the LCV membrane are not understood. Using cryo-focused ion beam milling, cryo-electron tomography, and confocal laser scanning fluorescence microscopy, we show that up to half of the intravacuolar L. pneumophila bacteria tether their cell pole to the LCV membrane. Tethering coincides with the presence and function of T4SSs and likely promotes the establishment of distinct contact sites between T4SSs and the LCV membrane. Contact sites are characterized by indentations in the limiting LCV membrane and localize juxtaposed to T4SS machineries. The data are in agreement with the notion that effector translocation occurs by close membrane contact rather than by an extended pilus. Our findings provide novel insights into the interactions of the L. pneumophila T4SS with the LCV membrane in situ.


RESULTS AND DISCUSSION
L. pneumophila tethers its cell pole to the LCV membrane. To understand ultrastructural features of the Icm/Dot T4SS during effector translocation, we infected two amoeba hosts-A. castellanii and D. discoideum-with L. pneumophila JR32 (a derivative of wild-type strain L. pneumophila Philadelphia-1, hereafter referred to as "wildtype") and analyzed intracellular bacteria using cryoFIB milling, cryoSEM, and cryoET. CryoFIB milling can reduce the thickness of voluminous biological samples, such as eukaryotic cells to a few hundred nanometers, allowing for subsequent high-resolution cryoET imaging of cell architecture as well as intracellular pathogens (41,42). Cryotomograms showed that shortly upon uptake into A. castellanii, L. pneumophila resided in a tight phagosome (Fig. 1A, 5 min postinfection [pi]), which was modified into the specialized and more spacious LCV as the infection progressed ( Fig. 1B and C, 30 min pi and 2 h pi). A distinguishing feature between phagosomes (Fig. 1A) and mature LCVs ( Fig. 1B and C) is that the latter are decorated with the endoplasmic FIG 1 L. pneumophila wild-type tethers its cell pole to the LCV membrane at early and late infection stages. A. castellanii amoebae were infected with L. pneumophila wild-type (A to C) or the DicmT mutant (E to G) for the time indicated. Representative 2D images of cryotomograms (12-nm tomographic slices) of L. pneumophila wild-type residing in a tight phagosome at very early infection stages (5 min pi, n vacuoles = 10) (A) or in a mature, ER-decorated LCV (30 min pi, n LCVs = 29; 2 h pi, n LCVs = 38) (B, C), with the cell pole tethered to the LCV membrane (;41% on average over time, n cell poles = 202). (D) Quantification of data shown in panels A to C and E to G. (E to G) T4SS-defective DicmT mutant bacteria oriented their poles to the vacuole membrane less frequently (;28% on average over time, n cell poles = 45). (H) Quantification of data shown in panels A to C and E to G. Data in panels D and H are represented as mean 6 standard deviation (SD) from at least two independent infection experiments (one-way ANOVA test; **, P , 0.01). OM, outer membrane; IM, inner membrane; Pha, phagosome; LCV, LCV membrane; Lcyt, L. pneumophila cytoplasm; Hcyt, host cell cytoplasm; ER, endoplasmic reticulum; mito, mitochondrion; asterisk, storage granule; Scale bars, 100 nm. reticulum (ER). Recruitment of the ER to the LCV is a hallmark of this replication-permissive pathogen compartment and is mediated by effector proteins such as SidC (43,44). While ER recruitment was not yet initiated at 5 min pi ( Fig. 1D; n vacuoles = 10), it became highly prominent at 30 min pi (65% of LCVs, n LCVs = 29), 2 h pi (49%, n LCVs = 38), and 3 h pi (46%, n LCVs = 38) in A. castellanii. Comparable results were obtained when D. discoideum was used as a host cell (see Fig. S1A to C in the supplemental material; 64% of LCVs at 0.5 h pi, n LCVs = 18; 43% of LCVs at 2 h pi, n LCVs = 23; 41% of LCVs at 3 h pi, n LCVs = 36). Data from both host cells therefore confirm that our infection protocols allow for the analysis of functional Icm/Dot T4SSs, which are required for LCV maturation and ER acquisition.
Interestingly, cryoET data also revealed that up to half of the bacterial poles remained in close proximity (,35 nm) to the LCV membrane. The poles indeed localized closer to the vacuole membrane than the lateral side of the bacteria throughout pathogen vacuole expansion and maturation ( To analyze whether the orientation of the bacterial poles toward the LCV membrane was dependent on the presence and/or function of the T4SS, we performed identical infection experiments with the L. pneumophila DicmT mutant strain, which lacks IcmT, an inner membrane component of the Icm/Dot T4SS. The DicmT mutant produces a defective T4SS, does not form a replication-permissive LCV, and shows severe defects in bacterial conjugation, macrophage killing, and survival in amoebae (10,45,46). Cryotomograms of A. castellanii (Fig. 1E to G; n = 18) and D. discoideum (Fig. S1C to E; n = 13) infected with the L. pneumophila DicmT strain revealed that these bacteria-containing vacuoles were never decorated with ER ( Fig. 1D), confirming the functional defect of the T4SS. Compared to the wild-type strain, cell pole tethering (defined as ,35-nm spacing between the bacterial outer membrane and the LCV membrane in a spacious vacuole) was observed at a significantly lower frequency for the DicmT mutant bacteria in A. castellanii (Fig. 1H) and D. discoideum (Fig. S1). Within our experimental time frame (3 h), we observed an ;1.5-fold reduction of cell pole tethering between the L. pneumophila wild-type and DicmT strain (on average 41% versus 28%; n cell poles-mutant = 45 and n cell poles-wild-type = 202). Taken together, these findings suggest that structural components of the Icm/Dot T4SS and/or its function (effector translocation) play a role in tethering the bacterial cell pole to the LCV membrane.
The Icm/Dot T4SS localizes to the poles of intracellular L. pneumophila. The Icm/Dot T4SS has been identified at bacterial cell poles in the L. pneumophila strain Lp02 grown in broth (16,36,37,40). We hypothesized that tethering of the bacterial cell pole to the LCV membrane might be a consequence of the polar localization of the T4SS. Accordingly, we analyzed by cryoET the localization of T4SSs in the L. pneumophila strain JR32 used in this study (strain JR32 and strain Lp02 are both Philadelphia-1 derivatives). Cryotomograms confirmed the polar localization of T4SSs in L. pneumophila JR32 as well as in DicmT mutant bacteria (see Fig. S3 in the supplemental material; average of ;3 T4SSs/cell pole, range of 1 to 6 T4SSs/cell pole, n cell poles = 47). While lateral T4SSs were identified at low frequency (;1.6 per cell) in a previous study (18), we did not observe nonpolar T4SSs in JR32 or the DicmT strain. This discrepancy could, however, be due to different strains and growth conditions used in the previous study (solid medium) and our study (liquid medium). Among the JR32 and DicmT mutant cell poles analyzed, 17% harbored one and 66% harbored multiple polar T4SSs ( Fig. S3; n cell poles JR32 = 28, n cell poles DicmT = 19). In conclusion, despite the functional defect of the L. pneumophila DicmT strain (Fig. 1E to G; see also Fig. S1) (10,45,46), the mutant bacteria retain the subunits required to assemble the characteristic "Wi-Fi-like" structure of the T4SS. Moreover, our data confirm previous reports on the polar localization and copy number of T4SSs in L. pneumophila (16,18,36,37,40).
We also assessed the localization of the Icm/Dot T4SS in amoebae infected with the L. pneumophila wild-type strain JR32. Cryotomograms of L. pneumophila inside LCVs in either A. castellanii or D. discoideum at 2 h pi revealed that the Icm/Dot T4SS localizes to the bacterial cell poles also intracellularly (see Fig. S4 in the supplemental material) and that the bacterial outer membrane closely associates with the LCV membrane (distance ,35 nm; n events = 41). Taken together, the Icm/Dot T4SS localizes to the L. pneumophila poles extracellularly as well as intracellularly.
Polar tethering of L. pneumophila correlates with Icm/Dot T4SS structure and function. To further investigate a potential correlation between polar tethering and the structure and/or function of the Icm/Dot T4SS, we quantified cell pole tethering of L. pneumophila wild-type and different icm mutant strains (DicmT, DicmE, DicmN, and DicmK strains). The Icm/Dot machinery is composed of 27 different subunits, and complexes lacking individual subunits might still assemble but adopt distinct, impaired structures (12). Indeed, while the DicmT, DicmE, and DicmN mutants lack structural components of the T4SS but still form complexes ( Fig. S3 and S5A in the supplemental material), the DicmK mutant (alias DdotH mutant) lacking the outer membrane component of the core transmembrane subcomplex does not assemble any T4SS complexes (36).
In order to analyze LCV tethering of several icm/dot mutant strains, we switched from the labor-intensive cryoFIB milling/cryoET approach to CLSM, since fluorescence microscopy is more amenable to a higher sample throughput. In parallel, we also switched from A. castellanii to D. discoideum because only D. discoideum is genetically tractable, and for this amoeba many fluorescent probes are available. Accordingly, phosphatidylinositol 4-phosphate [PtdIns(4)P]-positive LCVs and phagosomes/endosomes can readily be visualized in D. discoideum. In order to visualize bacteria-containing vacuoles, D. discoideum dually labeled with the LCV/PtdIns(4)P probe P4C-mCherry (47) and the endosomal marker AmtA-green fluorescent protein (GFP) (48) were infected with L. pneumophila wild-type or icm mutant strains and analyzed by CLSM ( Fig. 2A). Specific contact sites between the bacterial pole and the LCV membrane were quantified only for "expanded" LCVs (not for "tight" vacuoles, where the bacterium is firmly wrapped by the host membrane). Bacteria perpendicular to the focal plane were not considered.
Using this approach, more than 50% of wild-type L. pneumophila strains in LCVs were oriented with their poles in proximity to the vacuole membrane at 2 h pi ( Fig. 2A and B; n events = 60) and 8 h pi ( Fig. S5B and C; n events = 60), indicating that cell pole tethering might be required during various intracellular infection stages. Compared to the JR32 wild-type strain, we observed an approximately 2-fold reduction of cell pole tethering for the L. pneumophila DicmT strain at 2 h pi ( Fig. 2A and B) and 8 h pi ( Fig. S5B and C). The same reduction was observed for the DicmE strain ( Fig. 2A and B). For the L. pneumophila DicmK strain, we observed a roughly 4-fold reduction ( Fig. 2A and B), while the DicmN strain displayed only moderately lower levels of cell pole tethering, despite the lack of some structural components ( Fig. 2A and B). Interestingly, LCVs harboring the DicmN strain accumulated the PtdIns(4)P probe P4C-mCherry, and thus, the mutant bacteria still produced a functional T4SS. This was not observed for the other icm mutant strains, which therefore produced a functionally impaired T4SS ( Fig. 2A).
Overall, tethering was significantly lower for the DicmT and DicmE mutant strains, which produce incomplete T4SS complexes, indicating that structural T4SS components are required to establish tethering. The additional 2-fold reduction of tethering observed for the DicmK mutant, which lacks the entire T4SS structure, further supports this hypothesis. Since tethering was not completely abolished for the L. pneumophila DicmK strain, it seems likely that a fraction of cell poles might also associate randomly with the vacuole membrane. Nevertheless, the fact that tethering of the L. pneumophila DicmN strain coincided with increased T4SS activity suggests that the activity also contributes to cell pole tethering. In summary, CLSM confirmed our previous observations from cryoSEM and cryoET regarding polar tethering of L. pneumophila in LCVs ( Fig. 1; see also Fig. S1 and S2). The fluorescence microscopy approach further indicated that cell pole tethering is promoted by both the presence and activity of the Icm/Dot T4SS.
Dynamics of bipolar tethering of L. pneumophila to the LCV membrane are T4SS dependent. In a complementary approach to correlate tethering of the L. pneumophila poles to the LCV membrane with the structure of the Icm/Dot T4SS, we used live-cell fluorescence microscopy. To this end, D. discoideum Ax3 producing P4C-mCherry and AmtA-GFP was infected for 2 h with mCerulean-producing L. pneumophila wild-type strain JR32, DicmT, or DicmK mutant bacteria, and the dynamics of the bacteria within their vacuoles were assessed for 60 s each ( Fig. 3A; see also Movie S1 to S3 in the supplemental material). During this period of time, L. pneumophila wild-type, DicmT, or DicmK mutant bacteria made contact in a bipolar manner to the vacuole membrane for an average of ;40 s, ;25 s, or ;20 s, respectively (Fig. 3B). Intriguingly, LCVs harboring the L. pneumophila wild-type strain tended to be larger (possibly promoting intravacuolar bacterial motility) and were decorated with PtdIns(4)P, while vacuoles harboring the DicmT or DicmK mutants were smaller (possibly restraining intravacuolar bacterial motility) and lacked PtdIns(4)P. Taken together, L. pneumophila wild-type bacteria make contact to the PtdIns(4)P-positive LCV membrane significantly Ax3 producing P4C-mCherry (pWS032) and AmtA-GFP (pHK121) and infected (MOI of 5; 2 h) with mCerulean-producing L. pneumophila wild-type or DicmT, DicmE, DicmN, or DicmK mutant bacteria harboring plasmid pNP99. Examples are shown for contact between bacteria and the LCV membrane (top, white arrowheads) or no contact (bottom, red arrowheads). Scale bars, 2 mm (overview) and 1 mm (insert). (B) Quantification of data shown in panel A (n events = 60). Data are represented as mean 6 SD from three biological replicates (one-way ANOVA test; *, P , 0.05; **, P , 0.01; ***, P , 0.001). longer than the DicmT or DicmK mutant bacteria residing in smaller, endosomal vacuoles. These results are in agreement with the notion that a structurally intact and fully functional T4SS promotes tethering to the pathogen vacuole membrane.
L. pneumophila causes an indentation in the LCV membrane juxtaposed to a polar T4SS. Next, we analyzed in detail the architecture of the interactions between bacterial cell poles and the LCV membrane in infected A. castellanii and D. discoideum amoebae. As a general pattern, the bacteria localized either close to (,35 nm) or at quite a large distance (.300 nm) from the LCV membrane (n = 120). Strikingly, when the bacterial cell pole was within ;35 nm reach of the LCV membrane at 2 h pi (n = 41), ;30% of the observed T4SSs appeared to form distinct interaction sites, characterized by indentations in the LCV membrane juxtaposed to the precise site of a T4SS at the bacterial pole ( Fig. 4A to C; n contact sites = 12). We also identified multiple T4SS-LCV contact sites at the same bacterial pole ( Fig. 4D; n = 3). We did not observe similar interaction sites between bacterial cell poles and the proximal LCV membranes in the absence of a T4SS. Importantly, membrane indentations with similar spacing and curvature were also not identified when the bacterial poles were at a greater distance to the LCV membrane. Finally, similar interaction sites were not observed for the L. pneumophila DicmT strain (Fig. S1D to E). Legionella-LCV Contact Sites ® The distance between the bacterial outer membrane and the LCV membrane at these contact sites was on average ;16 nm ( Fig. 4; quantification area indicated in Fig. 4C as yellow area inside the box) compared to ;31 nm in close proximity to these contact sites (quantification area indicated in Fig. 4C as brown area inside the box). Overall, data collected from infection experiments using two different amoeba hosts did not indicate the presence of an extended potential T4SS pilus that would mediate effector translocation. Hence, effector translocation might occur either through a short (few-nm-long) pilus-like structure or through direct contact between the bacterial outer membrane and the LCV membrane, mediated by distal parts of the T4SS complex. Interestingly, at one L. pneumophila-LCV contact site, the bacterial outer membrane and the LCV membrane appeared somewhat "smeared," in agreement with the notion of membrane alterations at the contact site (Fig. 4B). In summary, our in situ data of L. pneumophila inside LCVs show tethering of bacterial cell poles to the LCV membrane during vacuole expansion and distinct contact sites ("indentations") between tethered bacterial poles and the LCV membrane, specifically where a functional T4SS is present.
LPS layers at Icm/Dot sites and L. pneumophila poles tethered to host membranes. To allow for secretion of effector proteins across the host cell plasma membrane and/ or the pathogen vacuole membrane, the pilus/conduit of a secretion system has to be long enough to cross the lipopolysaccharide (LPS) layer of the Gram-negative donor bacterium and establish contact with the target membrane. For instance, the length of the T3SS needle is actively regulated by a tape measure protein in various pathogens (e.g., Yersinia, Shigella, and Salmonella) to ensure that contact with the target membrane can be established (49)(50)(51)(52)(53)(54). As a result, Salmonella enterica serovar Typhimurium (S. Typhimurium) with longer LPS (ranging from 35 to more than 100 Oantigen repeat units) is severely impaired for invasion, as the T3SS needle cannot contact the host cell membrane anymore (50).
We sought to study by cryoFIB milling and cryoEM the bacterial LPS layer at Icm/ Dot sites and the L. pneumophila poles tethered to host membranes. To this end, we used human HeLa cells, which are better suited than amoebae for an analysis of early events during pathogen-host cell encounters. Specifically, we chose HeLa cells as a model, since their thin edges are easily accessible for cryoET imaging (55), and they have been shown to be more robust against high bacterial loads (56) than macrophages (57,58). HeLa cells are infected by L. pneumophila rather inefficiently, and some bacteria remain adherent to the cell surface (59); this allows to frequently capture and visualize bacteria in close association with the plasma membrane.
To estimate the minimal required length of a potential T4SS pilus/conduit, we quantified the thickness of the LPS layer in cryotomograms of L. pneumophila. To this end, we segmented cellular boundaries in cryotomograms, converted them to binary masks, and computed the minimal distance between the two masks as an estimate for the thickness of the LPS layer (see Fig. S6 in the supplemental material). Cryotomograms of L. pneumophila showed a contiguous electron-transparent layer around the bacterial outer membrane (Fig. 5A). The observed layer exhibited a reduction in density compared to that of the surrounding amorphous ice. Based on the thickness of this layer and previous reports from other bacterial species (60-62), we hypothesized that regions with reduced density around L. pneumophila reflect the presence of LPS.
Next, we analyzed the thickness of LPS layers during host cell entry. Using the quantification approach described above (Fig. S6), we found that LPS layers of planktonic L. pneumophila and bacteria adhering to HeLa cells (labeled as "invading") displayed an average thickness of 25.5 nm 6 2.2 nm (Fig. 5A and B), indicating that LPS forms a constant physical barrier between the bacterium and the host cell. Taken together, planktonic and adherent L. pneumophila strains show a similar LPS layer thickness.
Intriguingly, at the assembly sites of T4SSs, the LPS layer of L. pneumophila entering HeLa cells seemed to be slightly thinner and spanned 19.2 nm 6 1.6 nm ( Fig. 5C; 5 min pi; n T4SSs = 11). If the LPS were of the same length at these sites, we would expect the LPS to push into the HeLa plasma membrane and create a small deformation, corresponding to the bulge in the bacterial outer membrane caused by a T4SS. However, such irregularities were not observed at the host cell plasma membrane, indicating that LPS can potentially be "squeezed" (or assembles differently) to reduce the distance to the host membrane. Alternatively, T4SSs could preferably assemble at bacterial pole sites with slightly shorter LPS to facilitate the interaction between the T4SS and the LCV membrane. Importantly, the thinner LPS layer (;19 nm) correlates with our previous observation of ; 16-nm gaps at contact sites between the bacterial outer membrane and the LCV membrane (Fig. 4) and possibly allows for closer interactions between the bacterium and the host cell. In summary, the quantification of L. pneumophila LPS thickness revealed that the LPS layer is similar for planktonic and adherent/ invading bacteria (;26 nm), but slightly thinner at assembly sites of T4SSs (;19 nm), corresponding approximately to the gap between the bacterial outer membrane and the LCV membrane at the contact sites. These observations are in agreement with the notion that LPS adopts a specific structure at these sites to accommodate a functional T4SS.
To validate our quantification procedure, we performed control experiments with S. Typhimurium. Cryotomograms of planktonic S. Typhimurium revealed that LPS forms a layer of constant thickness around the bacterial cell body of ;30 nm (see Fig. S7A in the supplemental material). The thickness of the LPS layer was comparable between planktonic and invading bacteria (Fig. S7B), similar to what we previously observed for L. pneumophila (Fig. 5A and B). These results suggest that LPS layer thickness is not regulated during invasion and predetermines the distance between the bacterial outer membrane and the host cell membrane. Previous studies have shown that S. Typhimurium adapts its LPS upon host cell invasion to evade immune recognition (51,63,64). Our results revealed that at 1 h pi, the thickness of the LPS layer of intracellular S. Typhimurium is indeed reduced by ;5 nm (Fig. S7C), validating our approach to quantify the LPS layer in cryotomograms.
As an additional control, we quantified the thickness of the lipooligosaccharide (LOS) layer in Campylobacter jejuni. LOS is a low molecular weight form of LPS and consists of a lipid A that is linked to a polysaccharide but lacks the O-specific polysaccharide chain commonly found in other Gram-negative bacteria (65). Indeed, we found that the LOS layer of intracellular C. jejuni was significantly shorter (;15 nm) than the LPS layer in S. Typhimurium (;25 nm), further validating the accuracy of our LPS quantification protocols (Fig. S7D). Taken together, the LPS of planktonic or invading S. Typhimurium is of similar thickness (;30 nm) and thinner than LPS or LOS of intracellular S. Typhimurium or C. jejuni, respectively.
Finally, we sought to validate the HeLa cells as a model for Icm/Dot-dependent polar tethering of L. pneumophila to the LCV membrane. To this end, HeLa cells producing the PtdIns(3)P probe eGFP-2ÂFYVE were infected with mCerulean-producing L. pneumophila wild-type or DicmT, DicmE, DicmN, and DicmK mutant bacteria, and immunostained for the early endosome antigen 1 (EEA1). Both endosomal markers localize to vacuoles harboring L. pneumophila only early during infection, but allow visualization of the vacuoles (Fig. 5D). Quantification of bacterial tethering to the vacuole membrane revealed that the contacts of the DicmT, DicmE, and DicmK mutant bacteria were approximately 2-fold less frequent, compared to those of the parental strain ( Fig. 5E; n events = 45). These findings are very similar to what we observed in amoebae (Fig. 2). In summary, HeLa cells infected with L. pneumophila are a valid model to assess the Icm/Dot-dependent polar tethering of the pathogen to the LCV membrane.
Conclusions. In conclusion, our in situ data show that the L. pneumophila wild-type and DicmT strain harbor polar Icm/Dot T4SSs (Fig. 6A). Intracellular L. pneumophila wild-type tethers its cell pole more frequently to the LCV membrane (Fig. 6B) compared to the DicmT, DicmE, DicmN, and DicmK mutant bacteria, suggesting that both T4SS structure and activity are required to establish tethering. Tethering of L. pneumophila wild-type bacteria establishes distinct contact sites characterized by indentations in the LCV membrane toward the T4SS (Fig. 6C). We hypothesize that the T4SS might tether the L. pneumophila cell pole to the LCV during vacuole expansion. Tethering brings the T4SS in close proximity of the LCV membrane (Fig. 6B), thereby facilitating subsequent effector translocation. The importance of tethering for intracellular survival and proliferation of L. pneumophila as well as the signals and/or T4SS structural components triggering tethering and effector translocation are yet to be discovered (Fig. 6C). Our study provides first insights into the interactions between the Icm/Dot T4SS and the LCV membrane in the in vivo context and suggests that translocation of effector proteins might be achieved either by close membrane contact independent of an extended pilus or by only a short (,20 nm) pilus/conduit.  Table 1. L. pneumophila strains were grown for 2 to 3 days on charcoal yeast extract (CYE) agar plates (66), buffered with N-(2-acetamido)-2-aminoethane sulfonic acid (ACES) at 37°C. Liquid cultures in ACES yeast extract (AYE) medium (67) were inoculated at an optical density at 600 nm (OD 600 ) of 0.1 and grown at 37°C for 21 h to an early stationary phase (2 Â 10 9 bacteria/ml). Chloramphenicol (Cam) (5 mg/ ml) was added for plasmid retention. S. enterica Typhimurium SL1344 (wild type) was grown in LB broth supplemented with 0.3 M NaCl and 50 mg/ml streptomycin (Str) (AppliChem) for 12 h at 37°C. These cultures were subcultured (1:20) in the same medium and incubated for 4 h at 37°C. C. jejuni strain 108 was grown in heart infusion (HI) broth (Sigma) at 160 rpm or on blood agar (BA) plates (BA base no. 2; Sigma) supplemented with Campylobacter selective supplements and defibrinated horse blood (Oxoid; lysed with 5% saponin). Cultures were grown under microaerophilic conditions (CampyGen sachet; Oxoid) in anaerobic jars.
A. castellanii strain 5a2 (ATCC PRA-228) and D. discoideum strain Ax3 were grown as adherent cultures at 27°C and 23°C, respectively. A. castellanii was grown in Trypticase soy yeast (TSY) extract broth (30 g/liter TSY, 10 g/liter yeast extract; Sigma-Aldrich) and passaged every 5 days. D. discoideum Ax3 amoebae were cultivated in HL-5 medium (ForMedium) at 23°C in the dark. Cells were maintained every 2 to 3 days by rinsing with fresh HL-5 and by transferring 1 to 2% of the volume to a new T75 flask containing 10 ml of medium. Cells were strictly maintained below 90% confluence.
Transformation of D. discoideum. Transformation of axenic D. discoideum Ax3 amoeba was performed as described (44,68). Briefly, D. discoideum was grown to approximately 80% confluence. The cells were collected in fresh HL-5 medium by centrifugation (450 Â g, 5 min) and subsequently washed once with 5 ml Sorensen phosphate-buffer (SorC; 2 mM Na 2 HPO 4 , 15 mM KH 2 PO 4 , 50 mM CaCl 2 Â 2H 2 O, pH 6.0, autoclaved and stored at 4°C) and once with 5 ml electroporation buffer (EB) (10 mM KH 2 PO 4 , 50 mM sucrose, pH 6.1, filter-sterilized and stored at 4°C). The washing buffer was replaced with 2 ml fresh EB, and the cells were resuspended with a 5-ml serological pipette. Eight hundred microliters of the cell suspension each was added to a 4-mm gap electroporation cuvette (Bio-Rad). Two micrograms of both vectors were simultaneously added to the cuvette. Electroporation was performed with 2 pulses of 1 ms and 1 MV separated by a 5-s gap. Directly after electroporation, the cells were transferred into a T75 flask containing 10 ml HL-5. Around 24 h after electroporation, the medium was replaced with fresh HL-5, and the required selection antibiotics were added (20 mg/ml Geneticin, 50 mg/ml hygromycin). The medium was changed 72 h later. Upon obvious appearance of several microcolonies (usually 6 to 7 days after transformation), the cells were dislodged into fresh medium and transferred to a new flask.  (18), might activate the T4SS for effector transport across the cytoplasmic membrane ("cytoM transport"; step 1) by creating a channel within the cytoplasmic complex. These conformational changes can already be observed in L. pneumophila grown to stationary phase (A). Various, yet unknown, signals could further activate the T4SS to transport substrates across the bacterial outer and vacuole membrane (step 2). LCV, Legionella-containing vacuole; cytoM, cytoplasmic membrane; VM, vacuole membrane; OM, outer membrane; IM, inner membrane.
Infection assays for fluorescence microscopy. D. discoideum producing the desired fluorescent probes were harvested from approximately 80% confluent cultures, seeded at 1 Â 10 5 cells/ml in 6-well plates (Corning) or 8-well m-slides (for live-cell experiments) (ibidi) and allowed to adhere and grow for 24 h. Infections (multiplicity of infection [MOI] of 5) were performed with early-stationary-phase cultures of the L. pneumophila wild-type (JR32) and DicmT, DicmE, DicmN, or DicmK mutant strains harboring pNP99 (mCerulean), diluted in HL-5 and synchronized by centrifugation (450 Â g, 10 min, room temperature [RT]) (69). Subsequently, infected cells were washed three times with HL-5 and incubated at 25°C for the time indicated. Finally, infected amoebae were recovered from the 6-well plates, fixed with 4% paraformaldehyde (PFA) for 30 min at room temperature, transferred to 8-well m-slides, and embedded under a layer of phosphate-buffered saline (PBS)/0.5% agarose before imaging. For live-cell experiments, infected amoebae were directly imaged in the 8-well m-slides after incubation.
HeLa CCL-2 cells were harvested from approximately 80% confluent cultures, seeded at 7.5 Â 10 4 cells/ ml in 24-well plates (Corning) containing sterile coverslips and allowed to adhere for 24 h. Subsequently, cells were transfected with the desired fluorescent probe using the Lipofectamine 3000 kit (Thermo Fisher Scientific) according to the manufacturer's protocol and incubated for a further 24 h at 37°C/5% CO 2 . Infections (MOI of 150) were performed with early stationary cultures of the L. pneumophila wild-type (JR32) and DicmT, DicmE, DicmN, or DicmK mutant strains harboring pNP99 (mCerulean), diluted in DMEM, and synchronized by centrifugation (450 Â g, 10 min, RT). Subsequently, the infected cells were washed four times with DMEM and incubated at 37°C/5% CO 2 for the time indicated. Next, HeLa cells were fixed with 4% PFA for 20 min at RT, permeabilized with PBS/0.25% Triton X-100 for 20 min, and blocked with PBS/1% bovine serum albumin (BSA) (1 h, RT). The cells were then incubated with a primary antibody against EEA1 (Abcam; ab2900), diluted 1:100 in blocking buffer for 2 h at RT, followed by an Alexa Fluor 594-coupled secondary antibody (Thermo Fisher Scientific; A21442) at a concentration of 1: 200 (1 h, RT). Finally, the coverslips were washed three times with PBS and mounted on glass slides using ProLong Diamond antifade mounting medium (Thermo Fisher Scientific) and imaged.
Confocal microscopy of infected cells. Confocal microscopy of infected fixed or live cells was performed as described (23,47,69,70) using a Leica TCS SP8 X CLSM with the following setup: white light laser, 442-nm diode, and HyD hybrid detectors used for each channel. Pictures were taken using a HC PL APO CS2 63Â/1.4 oil objective with Leica Type F immersion oil and analyzed with Leica LAS X software. Settings for fluorescence imaging were as following: mCerulean (excitation, 442 nm; emission, 469 nm), eGFP (excitation, 488 nm; emission, 516 nm), mCherry (excitation, 568 nm; emission, 610 nm), and Alexa Fluor 594 (excitation, 590 nm; emission, 617 nm). Images and movies were captured with a pinhole of 1.19 Airy units (AU) and with a pixel/voxel size at or close to the instrument's Nyquist criterion of approximately 43 Â 43 Â 130 nm (xyz). A scanning speed of 400 Hz, bidirectional scan, and line accumulation equal to 2 were used to capture still images ( Fig. 2 and 5). Scanning speed of 700 Hz and bidirectional scan were used to capture movies (see Movies S1 to S3 in the supplemental material; Fig. 3).
Image processing. All images and movies were deconvolved with Huygens professional version 19.10 (Scientific Volume Imaging, The Netherlands; http://svi.nl) using the CMLE algorithm with 40 iterations and a 0.05 quality threshold. Signal to noise ratios were estimated from the photons counted for a given image. Single images, Z-stacks, and movies were finalized and exported with Imaris 9.5.0 software (Bitplane, Switzerland). Infection assays for electron microscopy. For electron microscopy, amoebae or HeLa cells were grown directly on grids as previously described (41). Briefly, amoebae (5 Â 10 5 per well) or HeLa cells (3 Â 10 4 per well) were seeded onto EM gold finder grids (Au NH2 R2/2, Quantifoil) and incubated for 1 h (amoebae) or overnight (HeLa) to allow the cells to attach to the grids. Infected T84 cells (1 Â 10 5 per grid) were directly applied onto the grids (Cu R2/1, Quantifoil). Cells were infected at an MOI of 75 (L. pneumophila for HeLa and A. castellanii), 100 (L. pneumophila for D. discoideum; C. jejuni for T84), or 300 (S. Typhimurium for HeLa). Samples were vitrified at different infection time points (L. pneumophila, 5 min, 30 min, 2 h, and 3 h; S. Typhimurium, 20 min and 1 h; C. jejuni, 2 h).
Cryo-focused ion beam milling. Cryo-focused ion beam (cryoFIB) milling was used to prepare samples of plunge-frozen infected amoebae for imaging by cryo-electron tomography (cryoET) (72). Frozen grids with infected cells were processed as previously described (41) using a Helios NanoLab 600i dual beam FIB/SEM instrument (Thermo Fisher). Briefly, lamellae with ;2 mm thickness were generated first (at 30 kV and ;400 pA). The thickness of the lamellae (final, ;200 nm) was then gradually reduced using decreasing ion beam currents (final, ;25 pA). Lastly, lamellae were examined at low voltage by cryoSEM imaging (3 kV, ;0.17 nA) to visualize intracellular bacteria. CryoFIB-processed grids were unloaded and stored in liquid nitrogen until further use.
Quantification of lipopolysaccharide and lipooligosaccharide layers. LPS and LOS layer quantification was done across the whole bacterial cell body, and accordingly, several hundreds of values were obtained for each condition. Cryotomograms were converted to TIFF images carrying the respective metadata. A representative 2D section was selected for segmentations of cellular boundaries (bacterium versus extracellular space; bacterium versus host cell) using the carving tool in ilastik (version 1.3.2 [74]) (see Fig. S6 in the supplemental material). Segmentations were exported as binary masks using ilastik. The masks were reduced to their edges (Canny edge detector) to compute the minimum distance from each pixel on one edge to the other edge. The gap between the two edges corresponds to the bacterial LPS or LOS layer. Data were processed and analyzed using Julia 1.0.3 and Images.jl 0.18.0 software.
Statistical analyses. Data from at least two independent infection experiments (for cryoET analysis) were analyzed by a two-tailed Student's t test to compare wild-type to mutant infections. Data from at least three ( Fig. 2 and 5) or two (Fig. 3) independent infection experiments (fluorescence microscopy) were analyzed by one-way analysis of variance (ANOVA).
Data availability. We declare that all data sets generated during this study are available from the corresponding authors upon reasonable request. Key tomograms have been submitted to EMDB (https://deposit-pdbe.wwpdb.org/deposition) under accession numbers EMD-13246, EMD-13247, EMD-13248, and EMD-13249.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. VIDEO S1, AVI file, 2.