Requirement of phosphatidic acid binding for distribution of the bacterial protein Lpg1137 targeting syntaxin 17

Many intracellular pathogens are known to manipulate a multitude of cellular functions to survive and grow inside the host. The intracellular pathogen Legionella pneumophila (L. pneumophila) is a well-studied bacterium that hijacks many kinds of cell types. It secretes more than 300 bacterial factors called Legionella effectors into the host through the bacterial type IV secretion system (Dot/Icm apparatus) to achieve hijacking (Hubber and Roy, 2010; Qiu and Luo, 2017). Although the exact functions of many Legionella effectors are still unclear, some of them have previously been identified as modulators of various cellular functions, such as membrane trafficking, autophagy, apoptosis, the unfolded protein response, and the signal transduction pathway (Hubber and Roy, 2010; Finsel and Hilbi, 2015; Qiu and Luo, 2017; Alshareef et al., 2021).

Upon host cell infection, several Legionella effectors localize to specific endomembrane compartments inside the host to exert their function, and they target these compartments by specifically binding to lipids such as phosphoinositides (PIs). For example, DrrA (encoded by the gene lpg2464), Lem4 (lpg1101), Lem28 (lpg2603), SidC (lpg2511) and SdcA (lpg2510) bind to phosphatidylinositol 4-phosphate (PI4P), and RavD, RavZ and RidL (encoded by lpg0160, lpg1683 and lpg2311, respectively) associate with phosphatidylinositol 3-phosphatase (PI3P) (Brombacher et al., 2009; Finsel et al., 2013; Hubber et al., 2014; Weber et al., 2014; Horenkamp et al., 2015; Romano-Moreno et al., 2017; Pike et al., 2019). Other effectors that potentially bind to PIs have been listed by Nachmias et al., and notably, these effectors have specific lipid-binding domains that are frequently located at their C-terminal regions (Nachmias et al., 2019).

Syntaxin 17 (Stx17) was originally identified as an endoplasmic reticulum (ER)-localizing SNARE protein in eukaryotic cells (Steegmaier et al., 2000), but recent studies have demonstrated that Stx17 localizes not only to the ER but also to mitochondria and the mitochondria-associated ER membrane (MAM) (Hamasaki et al., 2013; Arasaki et al., 2015). In addition to localizing to these various intracellular compartments, Stx17 is unique in that it is implicated in several important cellular functions. In nutrient-containing conditions (fed conditions), Stx17 promotes mitochondrial division through interaction with dynamin-related protein 1 (Drp1; also known as DNM1L), a GTPase implicated in mitochondrial fission on the MAM (Arasaki et al., 2015). Upon starvation, Stx17 dissociates from Drp1 and interacts with Atg14L (also known as ATG14), a component of phosphoinositide 3-kinase that participates in autophagosome formation, and this interaction leads to redistribution of Atg14L to the MAM at the site of autophagosome formation (Hamasaki et al., 2013; Arasaki et al., 2015). Furthermore, in the late stage of autophagy, autophagosomal Stx17 mediates membrane fusion between autophagosomes and lysosomes in coordination with vesicle-associated membrane protein 8 (VAMP8) and synaptosomal-associated protein 29 (SNAP29) (Itakura et al., 2012).

We previously identified a Legionella effector, Lpg1137, as a serine protease responsible for the degradation of Stx17 (Arasaki et al., 2017). This Stx17 degradation blocks not only autophagy, as expected, but also Bax-dependent apoptosis (Arasaki et al., 2017), a cellular defense system to prevent multiplication of pathogens. Subcellular fractionation analyses revealed that Lpg1137 associates with the ER, MAM, and mitochondria even though the cells lose Stx17 (Arasaki et al., 2017), implying that it may recognize these compartments by itself.

In this report, we found that the 60-amino-acid region at the C-terminal of Lpg1137 is sufficient for the intracellular distribution of Lpg1137, and this region preferentially binds to phosphatidic acid (PA). We also found that two basic amino acids (lysine and arginine) in this region are necessary for binding to PA. Finally, we demonstrated that an Lpg1137 mutant that is unable to bind to PA fails to localize to the target compartments and to degrade Stx17. Thus, we show how Lpg1137 targets the multi-distributional protein Stx17.

As it has been shown that some membrane-localizing Legionella effectors have lipid-binding domains in their C-terminal regions (Nachmias et al., 2019), we speculated that the endomembrane localization of Lpg1137 might be defined by its C-terminal region. To address this possibility, we constructed vectors encoding GFP-fused C-terminal regions of Lpg1137 (Fig. 1A) and expressed these constructs in HeLa-FcγRII cells. As shown in Fig. 1B, the distribution of the 30-amino-acid C-terminal region of Lpg1137 (GFP–Lpg1137 C30) was observed in a diffused pattern with some accumulation in the nucleus (top right panel), as with the distribution of GFP (top left panel). On the other hand, the fusion proteins containing the last 60 or 90 amino acids of the C-terminal region (C60 or C90) localized in a pattern resembling the endomembrane system (Fig. 1B, bottom panels), similar to that of full-length Lpg1137 (Fig. S1A). Fractionation analyses demonstrated that GFP–Lpg1137 C60 localized to the microsome, MAM and mitochondria (Fig. 1C). These results suggest that the endomembrane distribution of Lpg1137 is defined by its 60-amino-acid C-terminal region. Consistent with this, an Lpg1137 mutant that lacks C60 (GFP–Lpg1137Δ60) was found to be mainly localized in the cytosol (Fig. S1A,B). We narrowed down the endomembrane targeting site to amino acid residues 263-292 of Lpg1137 (GFP–Lpg1137 C60-30) (Fig. S1C,D). GFP–Lpg1137 C60 did not co-precipitate with FLAG–Stx17, which suggests that Lpg1137 distribution does not depend on its binding to Stx17 (Fig. S1E).

Given that the membranous localization of Lpg1137 does not appear to depend on the interaction with Stx17, we thought that Lpg1137 may bind to lipid(s) which exist in the ER, MAM, or mitochondria. To test this, we incubated maltose-binding protein (MBP) itself or MBP-fused Lpg1137 C60 with a nitrocellulose membrane in which 15 different lipids had been spotted, and then analyzed the lipid-binding specificity MBP–Lpg1137 C60 by immunoblotting using an anti-MBP antibody. As shown in Fig. 1D, a strong signal for MBP–Lpg1137 C60 was detected on a PA spot with weaker signals on cardiolipin (CL) and PI4P spots. PA is thought to be relatively enriched in the MAM and mitochondria, as PA is known to be a precursor of the mitochondrial-enriched phospholipid CL, and an ER-derived PA is transported to mitochondria through the MAM (Connerth et al., 2012; Potting et al., 2013; Tatsuta et al., 2014). Furthermore, Martinou and colleagues have previously shown that Drp1, which functions at the MAM and mitochondria, exhibits a similar lipid-binding pattern {binding preference of CL>PA>phosphatidylserine (PS)> phosphatidylinositol 4,5-phosphate [PI(4,5)P₂]} (Bustillo-Zabalbeitia et al., 2014). Therefore, our result that Lpg1137 binds to PA and CL seems reasonable. To gain more evidence that Lpg1137 directly binds to PA, we next employed a liposome floatation assay as depicted in Fig. 1E. When liposomes were prepared from dioleoylphosphatidylcholine (DOPC) and dioleoylphosphatidylethanolamine (DOPE), we could not detect the signal of MBP–Lpg1137 C60 in the top two fractions harboring floating liposomes (Fig. 1F, top panel). On the other hand, MBP–Lpg1137 C60 was detected in the top two fractions when liposomes contained dipalmitoylphosphatidic acid (DPPA) in addition to DOPC and DOPE (Fig. 1F, bottom panel). Interestingly, MBP–Lpg1137 C60 preferentially bound to saturated acyl chain-containing PA (DPPA) compared to the unsaturated form (dioleoylphosphatidic acid, DOPA) (Fig. 1G), as is the case with binding of Drp1 to PA (Adachi et al., 2016), suggesting that Lpg1137 C60 recognizes PA in a manner similar to that of Drp1.

Previous and recent works have revealed that Drp1 binds to PA and CL through its basic amino acids (Bustillo-Zabalbeitia et al., 2014; Adachi et al., 2016). As shown in Fig. 2A, there are two basic amino acid residues (lysine 266 and arginine 276) in the residues 263-292, the minimal region required for endomembrane distribution of Lpg1137. Therefore, we mutated each or both of them to alanine residues and observed the localization of these mutants. When expressed in HeLa-FcγRII cells, there were no significant differences in localization between GFP–Lpg1137 C60_{wild-type (WT)}, C60_K266A and C60_R276A (Fig. 2B, top two and bottom left panels). In contrast, expression of the double K226A and R276A mutant (GFP–Lpg1137 C60_K266/R276A) was observed in a cytosolic diffused pattern (Fig. 2B, bottom right panel). Consistent with this, most of the expressed GFP–Lpg1137 C60_K266/R276A was detected in the cytosolic fraction upon centrifugation (Fig. 2C, right panel). Furthermore, mutation of these residues caused a drastic reduction of affinity of Lpg1137 C60 for PA-containing liposomes (Fig. 2D), suggesting that Lpg1137 C60 targets endomembrane compartments to bind to PA through its two basic amino acid residues. We also demonstrated that these residues are important for the binding to CL (Fig. 2E).

We next examined whether mutation of the two basic amino acids affects the intracellular distribution and/or function of Lpg1137. To test this, we first assessed the effects of the mutations on endomembrane association of full-length Lpg1137. In GFP–Lpg1137_WT, almost equal amounts of the expressed protein were detected in the cytosolic and membrane fractions (Fig. 3A, first and second lanes from the left). On the other hand, the amount of the GFP–Lpg1137_K266/R276A in the membrane fraction was markedly reduced (Fig. 3A, rightmost lane). Consistent with this, the ratio of colocalization between GFP–Lpg1137_S68A (a protease-dead mutant that can associate with Stx17; Arasaki et al., 2017) and Stx17 was significantly reduced by additional K266/R276A mutations in GFP–1137_S68A (Fig. S2A). These results suggest that the two basic amino acids necessary for binding to PA define the intracellular distribution of Lpg1137.

Importantly, the association and proximity between Lpg1137_S68A and Stx17 in cells were abolished by the K266/R276A mutations (Fig. 3B; Fig. S2B). Consistent with this, degradation of Stx17 was not detected in cells expressing GFP–Lpg1137_K266/R276A (Fig. 3C,D) even though the purified mutant protein was found to retain protease activity toward purified His–Stx17 (Fig. S2C). These results suggest that targeting of Lpg1137 to the ER–mitochondria interface is necessary for binding to and degradation of Stx17 in cells.

We have previously shown that expressed Lpg1137 blocks Stx17-mediated autophagosome formation (Arasaki et al., 2017), and, thus, we next examined whether the double Lpg1137 mutant blocks autophagosome formation. Autophagosome formation, which was monitored by the formation of puncta marked by the LC3B (MAP1LC3B), was suppressed in starved cells expressing GFP–Lgp1137_WT (Fig. 4A, third row), whereas a significant number of LC3 puncta was observed in starved cells expressing GFP–Lpg1137_K266/R276A (Fig. 4A, bottom row). Consistent with this, LC3 lipidation (conversion of LC3-I to LC3-II upon starvation) and degradation of the autophagic substrate p62 (also known as SQSTM1) did not occur in Lpg1137-expressing cells, but occurred in Lpg1137_K266/R276A-expressing cells (Fig. 4B, comparison between lane 2 and 4 from left).

Finally, using a L. pneumophila mutant, we verified the importance of endomembrane localization of Lpg1137 for Stx17 degradation during infection. We cultured a FLAG–Lpg1137_WT- and FLAG–Lpg1137_K266/R276A-expressing L. pneumophila strain whose endogenous Lpg1137 had been depleted (Δlpg1137 strain) and confirmed the expression and translocation of FLAG–Lpg1137_WT and FLAG–Lpg1137_K266/R276A (Fig. S3A,B). As reported previously (Arasaki et al., 2017), degradation of Stx17 by Legionella infection was markedly suppressed in cells infected with the Δlpg1137 strain (Fig. 5A, comparison between lane 2 and 3 from left). Notably, degradation of Stx17 was restored in cells infected with the Δlpg1137 strain complemented with 3xFLAG–Lpg1137_WT, but not 3xFLAG–Lpg1137_K266/R276A (Fig. 5A, comparison between lane 5 and 6 from left). Consistent with this, the immunofluorescence intensity of Stx17 was drastically reduced in cells infected with the Δlpg1137 strain complemented with Lpg1137_WT, but not Lpg1137_K266/R276A (Fig. 5B, second column from the right), implying that membrane association of Lpg1137 through binding to PA is necessary for the recognition and degradation of Stx17 inside the host cell.

In this study, we provide a mechanistic insight into the endomembrane localization of Lpg1137, a Legionella effector with a serine protease activity toward Stx17. We show that the C-terminal 60-amino-acid region of Lpg1137 defines the localization of Lpg1137 to the ER, MAM and mitochondria through binding to PA, where Stx17 is predominantly distributed. Furthermore, we identified key amino acid residues of Lpg1137 for binding to PA, and mutation analyses demonstrated that the binding of Lpg1137 to PA is required not only for its endomembrane distribution but also for Stx17 binding and degradation in host cells.

It is striking that this is the first example of a Legionella effector that binds to PA. After internalization into the host, L. pneumophila creates a specialized replicative niche called the Legionella-containing vacuole (LCV), and PIs such as PI3P and PI4P are enriched in the LCV membrane (Haneburger and Hilbi, 2013). Therefore, most membrane-bound Legionella effectors which have functions on the LCV bind to PIs. Although RavZ, a Legionella effector, localizes to the isolation membrane but not the LCV membrane to suppress autophagy by cleaving phosphatidylethanolamine (PE)-conjugated LC3-II, it also recognizes one of the PIs, PI3P, which is accumulated in the isolation membrane (Choy et al., 2012; Horenkamp et al., 2015). In contrast, Lpg1137 specifically binds to PA. Although the lipid composition of the MAM has not been fully characterized, the MAM is thought to be enriched in phosphatidylserine (PS) and PA because these lipids are synthesized and dynamically transported from the MAM to mitochondria (Vance, 2014). Given that Stx17 is localized in the ER, MAM and mitochondria, the most efficient strategy for targeting Stx17 is that L. pneumophila effector(s) recognize lipid(s) that exist on the Stx17-associated compartments. This might be the reason why L. pneumophila bestows Lpg1137 with PA-binding ability. In addition, our previous work demonstrated that lysine 254 of Stx17 plays a crucial role in its distribution, and that Lpg1137 fails to degrade a lysine 254 mutant of Stx17 (Arasaki et al., 2015, 2017), raising the possibility that Stx17 may bind to PA through lysine 254 for its proper distribution. This possibility should be explored in the future.

Another important finding of this work is that Lpg1137 binds to PA in a manner similar to Drp1. In 2011, Voeltz and colleagues revealed that mitochondrial constriction and fission occurs at the ER–mitochondria contact sites and the mitochondrial fission factor Drp1 accumulates on these sites for promoting mitochondrial division (Friedman et al., 2011), suggesting that Drp1 localizes to the MAM as well as mitochondria.

Endocytic dynamin directly binds to PI(4,5)P₂ for the association with its target compartment which is the plasma membrane, and PA supports selective and deep insertion of dynamin into lipids (Burger et al., 2000). Recent studies have shown that Drp1 binds to PA and CL (Bustillo-Zabalbeitia et al., 2014; Adachi et al., 2016) and that the mitochondrial fission activity of Drp1 is controlled by its lipid-binding state: it is activated upon binding to CL and inactivated upon binding to PA (Adachi et al., 2016). Of note, Drp1 preferentially binds to PA (or CL) with saturated acyl chains rather than unsaturated forms (Adachi et al., 2016), suggesting that saturated PA is likely accumulated on membranes where Drp1 is localized, such as the MAM or mitochondria. More recently, Sesaki and colleagues discovered a novel role of Drp1; namely that it regulates the tubulation of the ER in a manner independent of mitochondrial fission activity, and they identified the minimal 18-amino-acid region required for the ER localization of Drp1 (Adachi et al., 2020). Notably, they also showed that a peptide consisting of these 18 amino acids directly binds to saturated PA, and that four lysine residues containing the peptide are essential for binding of the peptide to PA (Adachi et al., 2020). Thus, the Drp1-like PA binding mode of Lpg1137 is compatible with our conclusion that Lpg1137 is able to localize to the ER, MAM and mitochondria via binding to PA.

Previous bioinformatic protein structural analyses predicted that the structure of Lpg1137 is similar to that of mitochondrial carrier proteins such as solute carrier family 25 (SLC25) (Gradowski and Pawłowski, 2017). Because most of the mitochondrial carrier proteins localize to the mitochondrial inner membrane and mediate the transfer of molecules between the intermembrane space and the matrix (Klingenberg, 1990), it is possible that Lpg1137 localizes to the mitochondrial inner membrane. If so, it cannot cleave Stx17 because Stx17 localizes to the surface of the MAM and mitochondria. However, some studies show that SLC25A46, a member of the SLC25 protein family, localizes to the mitochondrial outer membrane (and perhaps MAM) and is involved in lipid exchange between the ER and mitochondria as well as maintenance of mitochondrial morphology (Abrams et al., 2015; Janer et al., 2016; Steffen et al., 2017), implying that some SLC25 proteins may have functions outside the mitochondria. Ugo1 (a yeast homologue of SLC25A46) also localizes to the mitochondrial outer membrane, and PA is required for the assembly of Ugo1 after association with the membrane (Vögtle et al., 2015). Interestingly, two basic amino acids (lysine 387 and lysine 407) of SLC25A46 are predicted to be located in an α-helix and might correspond to lysine 267 and arginine 276 of Lpg1137 (Fig. S3C), raising the possibility that SLC25A46 might bind to PA or CL via these residues. Further studies are needed to address this possibility.

Mouse monoclonal antibodies against calnexin (CNX; 1:1000 for immunoblotting, 610523) and Tom20 (1:300 for immunofluorescence, 1:1000 for immunoblotting, 612278) were purchased from BD Biosciences Pharmingen. Rabbit polyclonal antibodies against α-tubulin (1:1000 for immunoblotting T6074), FLAG (1:3000 for immunoblotting, F7425), and mouse monoclonal antibody against FLAG (1:300 for proximity ligation assay, F3265) were purchased from Sigma-Aldrich. Mouse monoclonal antibodies against MBP (1:1000 for immunoblotting, E8032), penta-His (1:1000 for immunoblotting, 34660), and rabbit polyclonal antibody against GFP (1:3000 for immunoblotting, 1:300 for proximity ligation assay, A6455) were purchased from New England Biolabs, Qiagen, and Thermo Fisher Scientific, respectively. Rabbit polyclonal antibodies against LC3B (1:150 for immunofluorescence, 1:1000 for immunoblotting, PM036), p62 (1:1000 for immunoblotting, PM045) and mouse monoclonal antibody against Stx17 (1:1000 for immunoblotting, M212-3) were obtained from MBL. Rabbit polyclonal antibody against Stx17 was prepared in our laboratory. Rabbit and mouse polyclonal antibodies against L. pneumophila were prepared as described previously (Arasaki et al., 2017).

GFP–Lpg1137 C30, C60, C90 and C60-30, or MBP–Lpg1137 C60 were constructed by inverse PCR using full-length GFP–Lpg1137 or MBP–Lpg1137 as a template, respectively. GFP–Lpg1137 C60_K266A, C60_R276A, C60_K266/R276A and GFP–Lpg1137_K266/R276A, or MBP–Lpg1137 C60_K266/R276A and MBP–Lpg1137_K266/R276A were also constructed by inverse PCR. The sequences of primers used are listed here. GFP–Lpg1137 C90, forward primer (Fw), 5′-AATAACCGTTTAATTACTGTG-3′, reverse primer (Rv), 5′-CTTGTACAGCTCGTCCATGCC-3′; GFP–Lpg1137 C60, Fw, 5′-ACCTTTTTCAAAGTTAGCTAT-3′, Rv, 5′-CTTGTACAGCTCGTCCATGCC-3′; GFP–Lpg1137 C30, Fw, 5′-TATATGGGTCCTCAACCTTTG-3′, Rv, 5′-CTTGTACAGCTCGTCCATGCC-3′; GFP–Lpg1137 ΔC60, Fw, 5′-GGTATGGATGAATAAATGGGTCCT-3′, Rv, 5′-TTACATGAAGGCTTCCTTAAGACCAAC-3′; GFP–Lpg1137 C60-30, Fw, 5′-GGTATGGATGAATAAATGGGTCCT-3′, Rv, 5′-TTATTCATCCATACCAAAAATGAGGGC-3′; MBP–Lpg1137 C60, Fw, 5′-ACCTTTTTCAAAGTTAGCTAT-3′, Rv, 5′-GAATTCTGAATCCTTCCCTCGAT-3′; Lpg1137_K266A mutation, Fw, 5′-ACCTTTTTCGCAGTTAGCTATTTG-3′, Rv, 5′-CATGAAGGCTTCCTTAAGACCAAC-3′; and Lpg1137_R276A mutation, Fw, 5′-CAAGTAGCGGTAGCAGCTCCTCAA-3′, Rv, 5′-CTGCAAATAGCTAACTTTGAAAAA-3′.

Maintenance of HeLa-FcγRII cells (parental HeLa cells; RIKEN, RCB0007) and HEK 293-FcγRII cells (provided by Dr Craig R. Roy, Yale University, USA) was performed as described previously (Arasaki and Roy, 2010; Arasaki et al., 2017). Transfection of plasmid DNA was performed using polyethylenimine (PEI; 24765-2; Polysciences) according to the manufacturer's recommendation. For starvation of cells, the cells were rinsed with PBS twice and then incubated in Earle's balanced salt solution (EBSS; Sigma Aldrich, E2888). L. pneumophila strains (Lp01 WT; CR39, Lp01 ΔdotA mutant; CR58; acquired from Dr Craig R. Roy) were grown on charcoal-yeast extract (CYE) plates as described previously (Roy and Isberg, 1997). The Lp01 Δlpg1137 mutant was established by allelic exchanges method using the pSR47S vector (Dr Ralph R. Isberg, Tuft University, USA) inserted into the 300-bp upstream and downstream regions of lpg1137 as described previously (Zuckman et al., 1999). To construct pMMB207-3x-FLAG–Lpg1137_WT or Lpg1137_K266/R276A, lpg1137 gene fragments including WT or K266/R276A mutant were amplified from the genomic DNA of the Lp01 strain by PCR. The amplified lgp1137 fragments were cloned into the pMMB207-3x-FLAG vector (Kubori et al., 2017). Finally, the resulting plasmids, pMMB207-3x-FLAG, pMMB207-3x-FLAG–Lpg1137_WT, or pMMB207-3x-FLAG–Lpg1137_K266/R276A were transformed into the Δlpg1137 strain. The sequences of primers used are listed here. Amplification of the lpg1137 gene: Fw, 5′-TGACAAGCGGATCCTCATGATTCAAAGAGGATTTACAATG-3′; Rv, 5′-CAAAACAGCCAAGCTTTTATTTCTTGGTAGGGGATG-3′. lpg1137_K266A mutation: Fw, 5′-TGACAAGCGGATCCTCATGATTCAAAGAGGATTTACAATG-3′; Rv, 5′-CTACTTGCTGCAAATAGCTAACTCTGAAAAAGGTCATGAAGGCTTC-3′. lpg1137_R276A mutation: Fw, 5′-CTATTTGCAGCAAGTAGCGGTAGCAGCTCCTCAAGCAGCCATCACTTTTG-3′; Rv, 5′-CAAAACAGCCAAGCTTTTATTTCTTGGTAGGGGATG-3′.

PIP strip (P-6002; Echelon) analyses were performed using 0.5 µg/ml MBP proteins according to the manufacturer's protocol. For the liposome floatation assays, lipids were purchased from Echelon (DOPC, L-1182; DOPE, L-2182; DPPA, L-4116; 16:0 CL, L-C160) and Avanti (DOPA, 840875P). The preparation of liposomes and liposome floatation assay were conducted as previously (Adachi et al., 2016) with slight modifications. In the step of mixing liposomes with proteins, 25 µg of proteins were used in each assay.

The five-point fractionation [postnuclear supernatant (PNS), cytosol, microsome, MAM, mitochondria] was performed as described previously (Arasaki et al., 2017). To perform the three-point fractionation (PNS, cytosol, total membrane), cells expressing GFP–Lgp1137 proteins were homogenized in buffer (0.25 M sucrose, 150 mM KCl, 20 mM HEPES-KOH, 2 mM EDTA, 1 µg/ml leupeptin, 1 µM pepstatin A, 2 µg/ml aprotinin, 1 mM PMSF and 1 mM dithiothreitol) by passaging five times through a 26G needle. The homogenates were centrifuged at 600 g for 5 min to obtain a postnuclear supernatant. The postnuclear supernatant was centrifuged at 100,000 g for 30 min using a S55A2 rotor (himac) to separate the cytosol and total membrane fractions. The protein concentration of each fraction was measured, and equal amounts of proteins were subjected to SDS-PAGE and then analyzed by immunoblotting.

Cells were fixed with 4% paraformaldehyde in PBS for 20 min at room temperature and permeabilized with 0.2% Triton X-100 for 15 min at room temperature. In infection experiments, the separation of extracellular and intracellular bacteria was performed as described previously (Arasaki et al., 2017). The samples were observed under a fluorescence microscope (Olympus BX50 or Keyence BZ-X800).

L. pneumophila was opsonized by rabbit or mouse anti-Legionella antibodies. Cells were spread at a density of 10⁶ cells and were infected with the L. pneumophila strains (WT, ΔdotA, Δlpg1137, Δlpg1137 3x-FLAG, Δlpg1137 3x-FLAG–Lpg1137_WT or Δlpg1137 3x-FLAG–Lgp1137_K266/R276A) for the indicated multiplicity of infection (MOI). At 1 h after infection, the infected cells were washed extensively and cultured in fresh Dulbecco's modified Eagle's medium (DMEM) for HEK293-FcγRII cells or fresh α-minimum essential medium (α-MEM) for HeLa-FcγRII cells for 3 h.

Purification of all recombinant proteins (MBP, MBP–Lpg1137 proteins, and His₆–Stx17) and in vitro degradation assay were performed as described previously (Arasaki et al., 2017).

HEK293-FcγRII cells expressing FLAG- and GFP-tagged proteins were lysed in lysis buffer containing 20 mM HEPES-KOH (pH 7.2), 150 mM KCl, 2 mM EDTA, 1 mM dithiothreitol, 1 μg ml⁻¹ leupeptin, 1 μM pepstatin A, 2 μg ml⁻¹ aprotinin, 1 mM PMSF and 1% Triton X-100. After centrifugation (17,400 g for 10 min), the supernatants were incubated with anti-FLAG M2 agarose (10 µl agarose beads; Sigma-Aldrich, A2220) for 1 h at 4°C. After incubation, the beads were washed extensively, and the precipitated proteins were eluted using SDS sample buffer and then analyzed by immunoblotting.

Proximity ligation assays (PLAs) were performed using PLA kits (Sigma-Aldrich) according to the manufacturer's protocol. The number of PLA dots were determined using the ImageJ software (NIH). Plasmids for GFP or GFP-tagged proteins (0.25 μg) were co-transfected with a plasmid for FLAG–Stx17 (1 μg). GFP or GFP-tagged proteins expressing cells were deemed to express FLAG–Stx17.

The results from each experiment were averaged and expressed as the mean with s.d. or s.e.m. and analyzed by a two-tailed paired Student's t-test (two groups) or a Tukey's test (more than three groups).

We thank Ms Ayaka Yanagase, Ms Yuma Sakaguchi and Ms Yuzuki Shimamori for their technical assistance.

The peer review history is available online at https://journals.biologists.com/jcs/article-lookup/doi/10.1242/jcs.259538.