Legionella effector Lpg1137 shuts down ER-mitochondria communication through cleavage of syntaxin 17

During infection of macrophages, the pathogenic bacterium Legionella pneumophila secretes effector proteins that induce the conversion of the plasma membrane-derived vacuole into an endoplasmic reticulum (ER)-like replicative vacuole. These ER-like vacuoles are ultimately fused with the ER, where the pathogen replicates. Here we show that the L. pneumophila effector Lpg1137 is a serine protease that targets the mitochondria and their associated membranes. Lpg1137 binds to and cleaves syntaxin 17, a soluble N-ethylmaleimide-sensitive factor attachment protein receptor (SNARE) protein that is known to participate in the regulation of mitochondrial dynamics through interaction with the mitochondrial fission factor Drp1 in fed cells and in autophagy through interaction with Atg14L and other SNAREs in starved cells. Cleavage of syntaxin 17 inhibits not only autophagy but also staurosporine-induced apoptosis occurring in a Bax, Drp1-dependent manner. Thus, L. pneumophila can shut down ER–mitochondria communication through cleavage of syntaxin 17.

L egionella pneumophila is a Gram-negative intracellular pathogen and is known to cause a serious type of pneumonia called Legionnaire's disease or a less serious one called Pontiac fever. In nature, L. pneumophila replicates inside protozoan hosts such as amoebae, but when inhaled by humans it enters alveolar macrophages and proliferates 1,2 . In macrophages, Legionella secretes effectors through the Dot/Icm apparatus that allow the conversion of the plasma membranederived vacuole to the Legionella-containing vacuole (LCV) by modulating membrane transport systems in host cells 3,4 . The LCV is not fused with lysosomes 5 nor undergoes autophagic clearance 6,7 , but intimately associates with vesicles derived from the endoplasmic reticulum (ER) 8 . ER-derived vesicles are recruited onto the LCV by the function of Legionella effectors such as DrrA/SidM, a guanine nucleotide exchange factor that activates Rab1 9,10 , tethered and fused with the LCV in a soluble N-ethylmaleimide-sensitive factor activating protein receptor (SNARE)-dependent manner 11,12 . As a result, the LCV membrane is converted from a plasma membrane-like structure into an ER/ER-Golgi-intermediate compartment-like structure, which is ultimately fused with the ER to create a compartment that supports bacterial replication 13 . At the initial stage of infection, mitochondria are found to be close to the LCV, whereas at the late stage they decline and ribosomes are observed around Legionella-containing membranes.
Syntaxin 17 (Stx17) is a SNARE originally implicated in a vesicle-trafficking step to the smooth-surfaced tubular ER membranes that are abundant in steroidogenic cells 14 . Stx17 is unique in that it has a long hairpin-like C-terminal hydrophobic domain (CHD), followed by a cytoplasmic basic region. Stx17 participates in cellular events unrelated to membrane fusion. In fed cells, Stx17 promotes mitochondrial fission by defining the localization and activity of the mitochondrial fission factor Drp1 (ref. 15). On starvation, on the other hand, Stx17 dissociates from Drp1 and associates with Atg14L, a subunit of the phosphatidylinositol 3-kinase complex. This promotes the recruitment of this kinase to the mitochondria-associated ER membrane (MAM) 16 , which leads to the formation of phosphatidylinositol 3-phophate (PI3P)-enriched omegasomes that are considered to represent a membrane/lipid source for autophagosomes 17,18 . In the late stage of autophagy, Stx17 present on autophagosomes mediates the fusion of autophagosomes with lysosomes [19][20][21] .
In this study, we show that Stx17 is degraded on Legionella infection. We identify the Legionella effector Lpg1137 as the responsible protein for Stx17 breakdown and show that Lpg1137 is a serine protease that localizes to the ER-mitochondria contact site, where Stx17 is located.

Results
Stx17 is degraded by the Legionella effector Lpg1137. In the course of a study to reveal the mechanism underlying Legionella infection and proliferation, we incidentally discovered that this pathogen cleaves Stx17 when engulfed by FcgRII-expressing HeLa cells (Fig. 1a,c) and macrophage-like THP1 cells (Fig. 1d). The cleavage of Stx17 by Legionella must be accomplished by its effectors because no cleavage was observed when cells were infected with an isogenic Legionella dotA mutant defective in effector secretion (Fig. 1c).
To determine which effector is responsible for Stx17 cleavage, we first used isogenic Legionella strains with large chromosomal deletions. L. pneumophila chromosome has a modular architecture, and five genomic islands (islands 2, 3, 4, 6 and 7) are devoid of genes for growth in rich media, but exhibit a high concentration of effectors that facilitate infection and proliferation in host cells 22 . Stx17 was found not to be cleaved in HeLa-FcgRII cells infected with the D2 mutant lacking 17 effectors, as well as the pentuple mutant (Fig. 1b,e). We constructed vectors encoding the individual effectors with a green fluorescent protein (GFP) tag and transfected them into HeLa-FcgRII cells. Stx17 staining was only lost in cells expressing the GFP version of the effector encoded by lpg1137 ( Supplementary Fig. 1). This was confirmed by immunoblotting (IB) (Fig. 2a) (Fig. 2b). Stx17 was not cleaved on infection with these mutants, as in the case of the dotA mutant (Fig. 2c). In contrast to the dotA mutant, on the other hand, these mutants did not exhibit a growth defect (Fig. 2d).
Lpg1137 is a serine protease localized in the MAM/mitochondria. To characterize Lpg1137, we first investigated its localization by subcellular fractionation. In addition to the cytosolic and microsomal fractions, GFP-Lpg1137 was recovered in the MAM and mitochondria fractions, where Stx17 resides (Fig. 3a). We therefore reasoned that Lpg1137 targets the MAM and mitochondria, and directly binds to and cleaves Stx17. To test this  idea, we transfected cells with a vector encoding GFP-Lpg1137 and then treated them with a serine protease inhibitor, phenylmethylsulfonyl fluoride (PMSF), or a proteasome inhibitor, MG132. We found that PMSF, but not MG132, substantially blocked Stx17 cleavage (Fig. 3b). Sequence analysis of Lpg1137 showed that Ser68 and the surrounding residues (-Gly-Leu-Ser-Gly-Gly-) fit the consensus sequence for the active site of serine proteases (Gly-X-Ser-X-Gly/Ala, where X is any residue).
As the S68A mutant lacks proteolytic activity, we could define the binding site for Lpg1137 on Stx17. The GFP-tagged S68A mutant bound to FLAG-Stx17, but barely to FLAG-Stx18 (Fig. 4a). Stx17 has a hairpin-like C-terminal hydrophobic domain comprising 44-amino-acid residues followed by the cytoplasmic C-terminal region (CHD þ C) that are required and sufficient for Stx17 to localize to the MAM/mitochondria and to interact with the mitochondrial fission factor Drp1 (ref. 15). Replacement of Lys254 with other amino acids except for Arg causes redistribution of Stx17 to the entire ER with loss of the Drp1-binding ability 15 . Analysis involving truncation mutants of Stx17 showed that the CHD þ C region, not the SNARE motif, is the region interacting with Lpg1137 (Fig. 4b). In addition, as in the case of the binding to Drp1, the K254C mutant exhibited reduced binding to Lpg1137, which is consistent with the fact that the K254C mutant was not cleaved on infection with Legionella (Fig. 4c).
Legionella eliminates the function of Stx17. We examined whether Legionella blocks the function of Stx17. The fact that Stx17 is cleaved by Lpg1137 predicts that the interactions of Stx17 with Atg14L in starved cells and Drp1 in fed cells are eliminated on Legionella infection. We investigated this prediction using  proximity ligation assay (PLA) 24 . When wild-type Legionella was used at a multiplicity of infection (MOI) of 50, almost all HeLa-FcgRII cells were infected (data not shown). Under this condition, PLA dots representing the proximity of Stx17 to 3 Â -FLAG-Atg14L in starved cells and to Drp1 in fed cells were largely diminished, whereas no significant decrease in the PLA signals was observed in cells infected with the dotA or Lpg1137 TM mutant ( Fig. 5a and Supplementary Fig. 2a). Moreover, the PLA signals for Stx17-3 Â -FLAG-Atg14L and Stx17-Drp1 were abolished on expression of GFP-Lpg1137 wild-type, but not GFP-Lpg1137 S68A ( Fig. 5b and Supplementary Fig. 2b).
Expression of Lpg1137 blocks autophagy. Given that Lpg1137 eliminated the Stx17-Atg14L interaction by cleaving Stx17, we next examined whether Lpg1137 could block autophagy. When cells expressing the Lpg1137 wild-type were starved for 2 h in the absence or presence of bafilomycin A1, the appearance of the autophagosome marker LC3 was reduced compared with in control cells (Fig. 6a). In contrast, no change in the formation of LC3 puncta was observed in cells expressing the S68A mutant (Fig. 6a). Inhibition of autophagy by Lpg1137 was confirmed by the reduced accumulation of LC3-II observed on IB (Fig. 6b). Autophagy starts with the formation of ER-connected omegasomes marked by the PI3P effector DFCP1 (refs 25,26). A previous study demonstrated that Stx17 recruits the PI3-kinase complex to the MAM through interaction with Atg14L 16 , a subunit of the kinase complex 27 . We therefore examined the localization of FLAG-DFCP1 and 3 Â -FLAG-Atg14L in GFP-Lgp1137-expessing cells after starvation. In GFP-expressing cells (control), both FLAG-DFCP1 and 3 Â -FLAG-Atg14L exhibited punctate staining, suggesting the formation of PI3P-positive omegasomes (Fig. 6c). In contrast, no discrete puncta for FLAG-DFCP1 or 3 Â -FLAG-Atg14L was observed in cells expressing the GFP-Lpg1137 wild-type. Expression of GFP-Lpg1137 S68A did not affect the formation of puncta for FLAG-DFCP1 and 3 Â -FLAG-Atg14L. These results suggest that GFP-Lpg1137 blocks an early stage of autophagy, that is, PI3P formation on omegasomes. Similar results were obtained when Stx17 was knocked down by short interfering RNA (siRNA) (Fig. 6d), corroborating that Stx17 participates in an early stage of autophagy. To further substantiate this view, we overexpressed RavZ, which inhibits autophagy by cleaving LC3 from preautophagosomal structures 6,7 . As reported, no LC3-positive structures were formed in GFP-RavZ-expressing cells, and the PI3P-binding protein GFP-RavZ was colocalized with mRFP-DFCP1 (Fig. 7a). Notably, FLAG-Stx17 was associated with GFP-RavZ-positive preautophagosomal structures. Moreover, depletion of Stx17 substantially suppressed the accumulation of GFP-RavZ as well as that of mRFP-DFCP1 on putative preautophagosomal membrane structures (Fig. 7b), consistent with the view that Stx17 regulates PI3P formation via recruiting the PI3-kinase complex 16 .
Legionella suppresses staurosporine-induced apoptosis. We explored the possibility that Stx17 cleavage might subvert cellular defense systems other than autophagy. Drp1, a dynamin-like GTPase whose activity is regulated by Stx17 (ref. 15), contributes to apoptosis by facilitating the recruitment of proapoptotic protein Bax to mitochondrial constriction sites 28,29 . In addition, a recent study showed that mitochondrial-anchored protein ligase  30 . We therefore examined the effect of GFP-Lpg1137 expression on apoptosis induced by the kinase inhibitor staurosporine (STS) and TNF-related apoptosis-inducing ligand ( TRAIL), which preferentially activate the proapoptotic proteins Bax and Bak, respectively 31 . In cells expressing the GFP-Lpg1137 wild-type, the degrees of nuclear condensation, caspase-3 activation and cytochrome c release from mitochondria induced by STS were markedly lower than those in GFP-or GFP-Lpg1137 S68A-expressing cells, whereas no difference was observed when apoptosis was induced by TRAIL ( Fig. 8a and Supplementary Fig. 3). Similar inhibition of STS-induced apoptosis was observed in cells infected by Legionella, but not by the dotA or Lpg1137 TM ( Supplementary  Fig. 4), and in cells depleted of Stx17 by siRNA ( Supplementary  Fig. 5). Drp1 knockdown also blocked apoptosis induced by STS, but not by TRAIL ( Supplementary Fig. 5). Notably, Bax failed to be translocated to mitochondria in cells depleted of Stx17 or Drp1 ( Fig. 8b and Supplementary Fig. 6), and therefore did not accumulate at mitochondrial constriction or fission sites (Fig. 8c) or at Drp1-positive structures (Fig. 8d) in cells depleted of Stx17.

Discussion
In the present study, we showed that Legionella Lpg1137 is a serine protease that targets the MAM/mitochondria, where it binds to and cleaves Stx17. No cleavage was observed for Sec22b or Atg14L, which are Stx17-binding partners in fed and starved cells [14][15][16] , respectively, suggesting the specificity of the Lpg1137 proteolytic activity. Cleavage of Stx17 blocks an early step of autophagy and STS-induced apoptosis. These effects were also observed in cells depleted of Stx17 by siRNA.
Studies on the function of Stx17 in autophagy showed, somewhat in contradiction, that this protein participates in both an early stage (autophagosome biogenesis) 16 and a late stage (autophagosome-lysosome fusion) [19][20][21] . The former is in line with the fact that mitochondria supply autophagosomes with lipids and proteins 32 . However, the interaction of Stx17 with Atg14L does not necessarily imply the involvement of Stx17 in the early stage of autophagy because Atg14L functions not only as a subunit of the PI3-kinase complex 27 but also controls Stx17mediated autophagosome fusion with lysosomes 21 . In the present study, using RavZ, a protease that cleaves the bond between LC3 and phosphatidylethanolamine 6,7 , we showed that Stx17 accumulates at areas containing PI3P in the absence of LC3. Moreover, depletion of Stx17 by overexpression of Lpg1137 or siRNA substantially reduced the formation of PI3P, as visualized by the absence of DFCP1-positive puncta. Thus, our data suggest that Stx17 participates in an early stage of autophagy. Lack of inhibition of autophagosome formation by silencing Stx17 observed in a previous study 19 might reflect the difference in Stx17 dependency of the early and late autophagy steps; autophagosome formation might be less sensitive to Stx17 depletion than autophagosome-lysosome fusion. The degree of Stx17 depletion in a previous study 19 might not have been enough for the inhibition of PI3-kinase complex recruitment step, but enough for the blockage of the fusion step.
In this study, we revealed a mechanism for the blockage of apoptosis by Legionella; cleavage of Stx17 by Lpg1137. For continued intracellular replication, Legionella needs to ensure the survival of the host cell against stresses by toxic microbial products, massive protein synthesis and the immune system. Legionella appears to use multiple mechanisms to prevent apoptosis. On infection of cells, the expression of antiapoptotic proteins is enhanced in host cells 33,34 . The Legionella effector SidF antagonizes proapoptotic proteins 35 . In addition, SdhA prevents apoptosis by an unknown mechanism 36 . A recent study reported that Legionella effectors such as Lgt1 and Lgt2 inhibit unfolded protein response by blocking the splicing of XBP1u mRNA, thus preventing host cells from unfolded protein response-driven apoptosis 37 .
In conclusion, the present observations in conjunction with the previous finding that Stx17 depletion impairs ER-mitochondrial Ca 2 þ homeostasis 15 suggest that Legionella can shut down the ER-mitochondria communication by cleaving Stx17. Given that the MAM and Stx17 are involved in a variety of cellular signalling and membrane trafficking pathways [14][15][16]19,38 , including the removal of oxidized or damaged mitochondrial content via mitochondria-derived vesicles 39 , it is tempting to speculate that cellular systems other than autophagy and apoptosis might be also subverted on Legionella infection.
In the course of this study, it was reported that the Legionella effector sphingosine-1 phosphate lyase (LpSpl) interferes with autophagy by affecting host sphingolipid metabolism 40 . Future work is required to reveal the interplay among multiple effectors (RavZ, Lpg1137 and LpSpl) to suppress autophagy, as well as the strategic advantage for the presence of the multiple autophagy suppressors in infection and intracellular growth of Legionella.

Methods
Animals. All animal procedures and experiments were approved by the Animal Care Committee of Tokyo University of Pharmacy and Life Sciences and conducted according to the guidelines of the committee.
Bacterial strains and plasmids. The conditions for the growth of L. pneumophila strains (the wild-type (Lp01), a thymidine auxotroph mutant (Lp02), a dotA mutant, a pentuple mutant and single island deletion mutants) and host cell infection were as described previously 6,22,[41][42][43] . The lpg1137::tn strain is in background SRS43 (Lp02::thyA (made by allelic exchange to incorporate the wildtype thyA gene)). The transposon mutants were made by electroporation of the plasmid pSRS_Cm1, which is a derivative of the pSAM_bt 44 with the original Erm resistance swapped for chloramphenicol resistance. In-frame deletion of the Lpg1137 gene was performed in Lp02 using allelic exchange, as described previously 43 . A cDNA fragment of Lpg1137 was generated by polymerase chain reaction (PCR) using primer pairs 1137 knockout (KO)-5F (5 0 -ATTGAGCTCTCC GAATCTTCATAATCTGG-3 0 ) and 1137KO-5R (5 0 -ATTGCGGCCGCTTGAAT which the deletion allele had replaced the wild-type allele on the chromosome, and loss of Lpg1137 was confirmed by IB using an antiserum against Lpg1137. Plasmid pcDNA-DEST-53 (Thermo Fisher Scientific) was used for the construction of all GFP fusion effectors encoded by the genes in island 2. Substitution of Ser-68 or Ser-134 was performed by inverse PCR. All FLAG-based Stx17 constructs were constructed as described previously 15 . 3 Â -FLAG-Atg14L and 3 Â -FLAG-DFCP1 (template for the FLAG-DFCP1 construct) were generous gifts from Dr Noboru Mizushima (University of Tokyo).
Cell culture and transfection. HeLa cells were obtained from the RIKEN Bioresource Center (RCB0007) and found to be free from mycoplasma contamination. HEK293 cells stably expressing FcgRII (HEK293-FcgRII cell) were maintained as described previously 11 . HeLa cells stably expressing FcgRII (HeLa-FcgRII cell) were established as described previously 11 , and grown in a-MEM supplemented with 50 IU ml À 1 penicillin, 50 mg ml À 1 streptomycin, 10% fetal calf serum, 2 mM L-glutamine and 400 mg ml À 1 hygromycin (Wako Chemicals). HeLa cells stably expressing wild-type FLAG-Stx17 or the K254C mutant were maintained as described previously 15 . THP1 monocytes were obtained from the RIKEN Bioresource Center (RCB1189) and grown in RPMI-1640 medium supplemented with 50 IU ml À 1 penicillin, 50 mg ml À 1 streptomycin and 10% fetal calf serum. THP1 monocytes were induced to differentiate into macrophages by treatment with 10 ng ml À 1 phorbol 12-myristate 13-acetate (Wako Chemicals) for 48 h. Transfection was carried out using Lipofectamine 2000 (Thermo Fisher Scientific) according to the manufacturer's protocol. (Sigma-Aldrich). The following antibodies were obtained from Sigma-Aldrich:  Protein purification and in vitro degradation assay. His 6 -tagged Stx17 fulllength was expressed in E. coli (BL21 codon-plus RP strain; Agilent Technologies) and solubilized in buffer containing 25 mM HEPES-KOH (pH 7.4), 500 mM NaCl, 1 mM MgCl 2 , 1 mM dithiothreitol, 1 mM PMSF and 1% of n-dodecyl-N,N-dimethylamine-N-oxide (Wako Chemicals). His 6 -Stx17 full-length was purified using Ni-NTA beads (Qiagen). MBP protein, MBP-tagged Lpg1137 wild-type and Lpg1137 S68A were expressed in E. coli (BL21 codon plus RP strain) and then solubilized in buffer containing 25 mM HEPES-KOH (pH 7.4), 500 mM NaCl, 1 mM MgCl 2 1 mM dithiothreitol and 1% of Triton X-100. The MBP proteins were purified using amylose resin (New England Biolabs). For in vitro degradation assay, 0.2 mg of Stx17 full-length was mixed with an equal amount of an MBP protein in reaction buffer containing 25 mM HEPES-KOH (pH 7.4), 50 mM NaCl and 5 mM MgCl 2 , followed by incubation at 37°C for 0, 30 or 60 min. The reaction was stopped by the addition of SDS sample buffer.
RNA interference. The siRNAs used were described previously 15 except for Atg14L (5 0 -UUUGCGUUCAGUUUCCUCACUGCGC-3 0 ). siRNAs were purchased from Japan Bio Services (Asaka, Japan). HeLa-FcgRII cells were grown on 35-mm dishes, and siRNAs were transfected at a final concentration of 200 nM using Oligofectamine (Thermo Fisher Scientific) according to the manufacturer's protocol.
Proximity ligation assays. PLA reagents were obtained from Sigma-Aldrich (catalogue numbers DUO92002, DUO92004 and DUO92008), and the assay was performed according to the manufacturer's protocol.
Induction of autophagy. To induce autophagy, cells were washed with PBS three times and then incubated with Earle's balanced salt solution (EBSS) for 2 h. To block autophagic flux, 0.1 mM bafilomycin A1 (Calbiochem) was included in EBSS.
Quantification of LC3-positive structures. As LC3-positive structures varied in size and large LC3-positive structures often aggregated in starved cells, we measured the LC3 immunostaining intensity, not the number of LC3-positvie dots, of each cell and compared. In this measurement, the intensity of LC3 signal was scored with the same threshold that eliminated almost cytosolic background of the LC3 signal.
Intracellular growth assay. Bone marrow macrophages were collected from C57BL/6 NLRC4 À / À mice 47 and plated in differentiation medium (RPMI supplemented with 20% fetal calf serum and 15% L929 supernatant). Differentiated bone marrow-derived macrophages were seeded at 2 Â 10 5 per well into 24-well tissue culture-treated plates one day before infection in replating medium (RPMI supplemented with 10% fetal calf serum and 7.5% L929 supernatant). Two-day patches of L. pneumophila strains were resuspended in sterile water and used to infect macrophages at an MOI of 1 in triplicates. Plates were centrifuged at 250 g to facilitate bacterial attachment. One hour postinfection, cells were washed three times in sterile PBS and 1 ml of replating medium was added back to the cells. Bacteria were liberated by hypotonic lysis at 1, 24, 48 and 72 h postinfection, and colony-forming units were enumerated on CYE agar after 4 days of growth at 37°C.
Quantification and statistics. To quantify nuclear condensation, cleavage of caspase-3 and cytochrome c release, at least 50 cells were counted in each experiment. Quantification of the fluorescence intensity of Bax, FLAG-DFCP1 and 3 Â -FLAG-Atg14L was performed using the ImageJ software. To quantify the number of PLA dots, 30 cells were scored in each experiment. The results were averaged, expressed, unless otherwise stated, as the mean with standard deviation and then analysed by means of a paired Student's t-test. The P values are indicated by asterisks in the figures with the following notations: *Po0.01, **Po0.001 and ***Po0.0001.