Bacterial secretion system skews the fate of Legionella-containing vacuoles towards LC3-associated phagocytosis

The evolutionarily conserved processes of endosome-lysosome maturation and macroautophagy are established mechanisms that limit survival of intracellular bacteria. Similarly, another emerging mechanism is LC3-associated phagocytosis (LAP). Here we report that an intracellular vacuolar pathogen, Legionella dumoffii, is specifically targeted by LAP over classical endocytic maturation and macroautophagy pathways. Upon infection, the majority of L. dumoffii resides in ER-like vacuoles and replicate within this niche, which involves inhibition of classical endosomal maturation. The establishment of the replicative niche requires the bacterial Dot/Icm type IV secretion system (T4SS). Intriguingly, the remaining subset of L. dumoffii transiently acquires LC3 to L. dumoffii-containing vacuoles in a Dot/Icm T4SS-dependent manner. The LC3-decorated vacuoles are bound by an apparently undamaged single membrane, and fail to associate with the molecules implicated in selective autophagy, such as ubiquitin or adaptors. The process requires toll-like receptor 2, Rubicon, diacylglycerol signaling and downstream NADPH oxidases, whereas ULK1 kinase is dispensable. Together, we have discovered an intracellular pathogen, the survival of which in infected cells is limited predominantly by LAP. The results suggest that L. dumoffii is a valuable model organism for examining the mechanistic details of LAP, particularly induced by bacterial infection.

also interferes with the classical endocytic pathway, we examined acquisition of the tethering protein early endosomal antigen (EEA) 1, and the lysosomal marker LAMP-1. These markers have previously been used to examine Dot/Icm-dependent trafficking of L. pneumophila 16,59 . EEA1 association was not commonly observed on LdCVs in THP-1 cells (Fig. 1A). To examine whether or not this is a Dot/Icm-dependent inhibition we created an L. dumoffii mutant lacking the dotA gene. The L. dumoffii ∆ dotA strain failed to prevent EEA1 association (Fig. 1A), and was unable to replicate inside host cells ( Figure S1A). Consistent with a block in endosomal maturation, LAMP1-association with LdCVs was rarely observed on LdCVs at one hour post-infection. However, we unexpectedly observed delayed acquisition of LAMP1 with ~10% of LdVCs positive at 3 to 4 hours post-infection ( Fig. 1B,C).
Taken together, L. dumoffii inhibits endosomal maturation and creates an ER-like niche, as previously described for L. pneumophila. However, the slow kinetics of maximal LAMP-1 acquisition was unexpected.
LC3 associates with a subset of LdCVs in a T4SS-dependent manner. In addition to the classical endosome-lysosome pathway, another process that could deliver bacterial pathogens to LAMP-1-positive compartments is autophagy. Thus, we examined whether LdCVs acquire the autophagy marker LC3. In RAW 264.7 cells stably expressing GFP-LC3, GFP-LC3 localization was clearly observed on a minority of LdCVs ( Fig. 2A). We then compared the LC3-recruitment phenotype of L. pneumophila and L. dumoffii in THP-1 cells (Fig. 2B). LpCVs rarely showed LC3 association, but LC3 was associated with up to ~20% of LdCVs. The peak of association was observed at three hours post-infection. LC3 was not observed on mature vacuoles containing replicating L. dumoffii (Fig. 2C). Importantly, we found that decoration with LC3 was Dot/Icm-dependent as our ∆ dotA deletion strain remained LC3-negative (Fig. 2B).
To examine whether LdCV LC3-acqusition involves host Atg machinery we assessed LC3 association to LdCVs in Atg7 +/− and Atg7 −/− mouse embryonic fibroblasts (MEFs). In the Atg7 −/− MEFs LC3-was not observed (Fig. 2D). Thus, this finding suggests that autophagy components engage with L. dumoffii in a process that requires the virulence activities of the pathogen.
Scientific RepoRts | 7:44795 | DOI: 10.1038/srep44795 LdCVs are devoid of autophagy signaling molecules and adaptors. We next examined the molecular mechanism behind LC3-decoration of L. dumoffii. Recognition of substrates decorated with ubiquitin, a 76-amino acid polypeptide, is a common feature of selective autophagy 61,62 , and, to date, all Legionella species tested have been shown to reside in vacuoles that are decorated with ubiquitin 63 . Thus, we analyzed whether L. dumoffii also resides within a ubiquitin-positive compartment that could contribute towards host recognition by autophagy. To do this, we infected THP-1 cells and assessed ubiquitin-localization using the ubiquitin (clone FK2) antibody, which detects both mono-and poly-ubiquitin conjugates but not free ubiquitin, at various stages of infection. Although LpCVs were decorated with ubiquitin, we did not detect significant localization at any stage of infection with LdCVs (Fig. 3A,B). We also examined ubiquitin-localization in cells co-infected with both L. pneumophila (mCherry/Hoechst 33342 double stained in Fig. 3C) and L. dumoffii (Hoechst 33342 single stained). Whereas ubiquitin-decoration was consistently observed for LpCVs, LdCVs found in the same cell remained ubiquitin-negative. This result further confirmed that LdCVs are largely devoid of ubiquitin. We also confirmed that LC3-positive LdCVs are largely ubiquitin-negative (Fig. 3D,F). Salmonella enterica serovar Typhimurium (S. Typhimurium)-containing vacuoles (SCVs), which are established for ubiquitin-mediated recognition by autophagy 64 , were found to be mostly ubiquitin-positive (~90%) (Fig. 3E,F). These results suggest that formation of LC3-positive LdCVs occurs via ubiquitin-independent mechanisms.
We next examined involvement of signaling molecules and adaptor proteins that are commonly found to target pathogens for LC3-decoration. Once again we used S. Typhimurium as a positive control because it is reported to recruit the 'danger' signal galectin-8 and the LC3-binding adaptors p62, NDP52, and Tecpr-1 on its vacuoles 38,39,42,43,65 . Consistently, localization of GFP-galectin-8, GFP-NDP52, GFP-p62 and Tecpr1-GFP was readily observed on SCVs (Figs 3G and S2). In contrast, localization of these constructs was rarely (< 2%) observed on LdCVs. Thus, it appears that decoration of L. dumoffii with LC3 is largely independent of ubiquitin, galectin-8, and the adapters NDP52, p62 and Tecpr-1.

LC3-associated LdCVs are single-membrane-bound vacuoles.
It is emerging that Atg proteins play roles in processes beyond classical autophagy 66 . Indeed, two degradative pathways are associated with the core autophagy marker LC3/Atg8: autophagy and LC3-associated phagocytosis (LAP). Because LAP is not associated with ubiquitinated proteins 47 and LdCVs lack ubiquitin and common autophagy adaptors, we decided to determine the pathway required for LC3-association. One of major differences between autophagy and LAP is the formation of a double-membrane-bound autophagosome during autophagy versus a single-membrane-bound LAP-phagosome. Thus, we examined the membranes of LC3-positive LdCVs using correlative light and electron microscopy (CLEM) at 2.5 to 3 hours post-infection (Fig. 4). This technique allows confocal and electron-micrographs to be taken of the same cell 57 . We examined 15 separate LC3-positive LdCVs from seven independent dishes. The vast majority of LC3-positive bacteria resided in spacious vacuoles with a single membrane ( Fig. 4A-C, LdCVs #1 and #2). Small intracellular vesicles were also observed within the lumen of these enlarged LC3-positive compartments (Fig. 4C, LdCV #1, white arrows). These may be artifacts or perhaps intraluminal vesicles 67 . LC3-negative vacuoles had attached vesicles (Fig. 4D, white arrows) and adjacent mitochondria    68,69 . Only a single double-membrane structure was found associated with an LC3-positive LdCV (Fig. 4A,B,E, #4). We also examined a rare LC3-positive L. pneumophila vacuole and observed a multi-membrane structure that is consistent with conventional autophagy (Fig. 4F). These results are consistent with the hypothesis in which a subset of LdCVs is targeted to LAP rather than conventional autophagy.
Ulk1 is dispensable for LC3-decoration of L. dumoffii. A mechanistic variable known to distinguish LAP from conventional autophagy is the requirement for the ULK (Ulk1-Atg13-Fip200) complex: Ulk1 (Unc51-like kinase autophagy activating kinase (1) is not required for LAP 50,74,75 . Conversely, conventional autophagy is Ulk1-dependent. Under non-inducing conditions this protein is functionally inhibited by the mammalian-target of rapamycin (mTOR), but after relief from repression the ULK complex activates formation of the initial isolation membrane 76 . We found that TLR2, but not TLR4, expression in non-immune HEK293 cells allows LdCV enhanced LC3-recruitment (compare control siRNA columns in Fig. 5D). The result on the transfectable cell-line enabled us to analyze the role of Ulk1 using siRNA knockdown. HEK293 cells were co-transfected with TLR2 and siRNA-targeting Ulk1 or scrambled siRNA. Levels of LC3-association with L. dumoffii were found to be comparable despite the reduced level of Ulk1 (Figs 5D and S4A,B). To validate the effect of Ulk1 knockdown, in uninfected cells we compared the effect of the autophagy activator rapamycin on LC3-puncta formation following scrambled or Ulk1 siRNA treatment. As expected, the addition of rapamycin to cells treated with scrambled siRNA led to increased LC3 puncta, but in Ulk1-knockdown cells LC3-puncta were L. dumoffii infection and LC3 detection was carried out as described for panels A-C. absent despite rapamycin treatment ( Figure S4B bottom panels). Thus, unlike conventional autophagy, the acquisition of LC3 on LdCVs is independent of Ulk1. Treatment with rapamycin has also been described to enhance levels of LC3 associated with pathogens, including S. Typhimurium 77 . However, treatment with this drug did not enhance LC3-decoration of L. dumoffii. Rather levels were slightly decreased ( Figure S4C).
Taken all together, we concluded that LC3-recruitment to a subset of Dot/Icm T4SS + LdCVs is mediated by LAP, a process independent of the mTOR-Ulk1 axis. Recently Rubicon was reported to be involved in LAP induction by fungal infection 78 . In BMDMs obtained from Rubicon −/− knockout mice, LC3-positive LdCVs were formed at a lower level than the Rubicon +/− control (Fig. 5E), providing further support for this notion. Further, unlike other pathogens that gain LC3 through multiple mechanisms, such as Salmonella, LAP appears to be the major mechanism targeting L. dumoffii for LC3-decoration.

Membrane damage is not associated with LAP induced by L. dumoffii infection. Classical
autophagy is often associated with membrane damage 25,79 , but whether membrane damage is important for LAP-activation is not clear. Galectin-3 has been used a marker for damaged phagosomes 80 . Thus, we examined the localization of galectin-3 on LdCVs as a measure of membrane integrity. As a control we again utilized S. Typhimurium, as galectin-3 has been found on SCVs 42 . Unlike SCVs, LdCVs were largely devoid of galectin-3 (Fig. 6A). To further examine the relationship between LAP-targeting of LdCVs and membrane damage, we next examined LC3-positive phagosomes for the presence of galectin-3. Whereas LC3-positive SCVs were commonly positive (~43%) for galectin-3, this marker was rare (< 5%) on LC3-positive LdCVs (Fig. 6B,C). Taken together, our data demonstrated that LAP-activation on LdCVs is independent of membrane damage.

LAP induced by L. dumoffii infection involves NADPH oxidases and diacylglycerol.
Nicotinamide adenine dinucleotide phosphate (NADPH) oxidases produce reactive oxygen species (ROS) in response to phagocytosis or inflammatory mediators, and this 'respiratory burst' acts directly to kill invading organisms 81 . Additionally, a role for the NADPH oxidases in recruitment of LC3 to phagosomes (Fig. 7A), including phagosomes harboring S. Typhimurium and L. monocytogenes, was demonstrated using the NADPH oxidase inhibitor diphenyleneiodonium (DPI) and Nox2 −/− mice 45,46 . Similarly, we found addition of 10 μ M DPI upon L. dumoffii infection drastically impaired LC3-acquisition (Fig. 7B). This suggests that NADPH oxidases are involved in activating LAP upon L. dumoffii infection.
In addition to other autophagic recognition mechanisms [38][39][40][41][42][43]82 , both S. Typhimurium and L. monocytogenes are decorated with diacylglycerol (DAG) during infection and this signal has been shown to be involved in LC3 association 46,47,83 . Manipulation of DAG levels using chemical inhibitors was found to perturb LC3-association with both organisms. Similarly, we found that specific chemical interventions that perturb DAG  (Fig. 7A), via inhibition of phospholipase D (PLD) by 1-butanol or phosphatidic acid phosphatase (PAP) by propranolol hydrochloride (Propr.), severely reduced association of LC3 with LdCVs (Fig. 7B). In contrast tert-butanol, an isomer of 1-butanol that has no inhibitory effect on PLD, had no effect on LC3 association with LdCVs. Normally, DAG kinase (DGK) reduces DAG levels by conversion of DAG to phosphatidic-acid. By inhibiting DGK with DGK inhibitor (R59022), which is expected to promote DAG accumulation, we reproducibly observed slightly but not significant enhanced levels of LC3-association with LdCVs (Fig. 7B). Furthermore, our data suggests both PAP and PLD function during L. dumoffii infection in a manner promoting LC3-association. The link between DAG and LC3 is thought to be PKCδ -binding via its C1-domain to vacuolar DAG, and subsequent activation of JNK and NADPH oxidases 47 . Indeed, inhibition of JNK activation using the drug SP600125 also reduced association of LC3 with LdCVs, though not as completely as inhibition of DAG production or NADPH oxidases (Fig. 7B). Taken together, our results support a role for both TLR2 and ROS generation via DAG-NADPH oxidases axis in triggering LAP.
LAP limits intracellular survival of L. dumoffii. To elucidate whether this host response limits L. dumoffii inside host cells, or whether L. dumoffii utilizes Atg-proteins for its intracellular survival strategy we first examined whether compartments positive for both LC3 and LAMP-1 exist. Classically, when LC3-positive autophagosomes fuse with lysosomes, autophagolysosome will be decorated with both LC3 and LAMP-1 before degradation of phosphatidylethanolamine (PE)-conjugated LC3 found within the lumen of the membrane-bound compartment. In RAW cells stably expressing GFP-LC3, LC3-positive LdCVs were also positive for LAMP1 by indirect immunofluorescence (Fig. 8A). In THP-1 cells, using dual staining, we found that at 3 hours post-infection ~50% of LC3-positive LdCVs were also positive for LAMP-1 (Fig. 8B,C). Thus, delayed acquisition of LAMP-1 of LdCVs is associated with LAP. As a control we also examined LC3-positive LdCVs for the presence of ubiquitin. LdCVs were rarely positive (~10%) for both LC3 and ubiquitin, consistently with our previous finding that only ~2% of total LdCVs are ubiquitin-positive (Fig. 3B).
To examine the fate of bacteria within LC3-positive LdCVs we then performed live imaging experiments. We did not observe LC3-positive LdCVs being expelled from the cells, as is the case for uropathogenic E. coli 67 , but rather the mCherry expressing bacteria within LC3-positive compartments disappeared over time ( Figure S5). Given that a number of pathogens interfere with host processes that acidify lysosomes to create a niche in which to replicate, we also confirmed that LdCVs become acidified by live imaging of RAW cells stably expressing GFP-LC3 infected with Hoescht-stained L. dumoffii strains in the presence of LysoTracker ® Red ( Figure S6). As a positive control we tracked ∆ dotA mutant L. dumoffii. ∆ dotA mutants were found within LC3-negative but acidic (lysotracker-positive) compartment shortly after infection, which is consistent with trafficking through the canonical endocytic pathway ( Figure S6A, top panels). Conversely, acidified L. dumoffii vacuoles were not observed at one hour post-infection ( Figure S6A, bottom panels). However, at 3 hours, LC3-positive but not acidified vacuoles were mainly observed and at 4-5 hours a positive lysotracker signal was associated with ~10% of L. dumoffii and LC3-positive LdCVs were rare ( Figure S6B). Thus, acidification occurs post acquisition of LC3. This data further supports degradation of pathogenic L. dumoffii that are recognized by Atg-components.
Because LAP of L. dumoffii results in bacterial degradation, we hypothesized that this innate immune response acts to restrict L. dumoffii. We examined this hypothesis using a bacterial viability-based assay. In wild-type MEF cells, ~25% more L. dumoffii remained viable in comparison with in Atg7 knockout MEF cells at five hours post infection (Fig. 8D). Likewise, in BMDMs derived from TLR2 −/− knockout mice, ~80% more L. dumoffii remained viable in comparison with in BMDMs derived from TLR2 +/− mice (Fig. 8E). Taken together, these results demonstrated that LAP can limit intracellular survival of a virulent bacterial pathogen that resides within a membrane-bound compartment within host cells.

RavZ is sufficient to block LAP induced by L. dumoffii infection. Some L. pneumophila strains
including Lp01 we used in this study encode the effector protein RavZ, which irreversibly deconjugates LC3 from phospholipid membrane 17 and thus interferes LAP 78 . To explore the exact role of RavZ in LAP avoidance of LpCVs, we examined ubiquitin and LC3 recruitment to LpCVs containing a RavZ detetion strain in THP-1 cells at three hours post infection, In this condition we observed LC3 recruitment to LdCVs but rarely to LpCVs (Fig. 9A,B) like LpCVs containing the wild-type strain (Fig. 3B). This clearly indicates that the LAP avoidance of L. pneumophila is not solely due to the effector RavZ function.
On the other hand, L. dumoffii NY23 used in this study does not encode RavZ ortholog. When L. pneumophila RavZ or its catalytic mutant derivative (RavZC258A) was ectopically expressed in L. dumoffii, the resulting LdCVs failed to acquire LC3 in a RavZ activity-dependent manner (Fig. 9C), suggesting that the catalytic activity of RavZ is sufficient to block LAP as previously reported 78 .

Discussion
To lay a foundation for studies examining species-specific Legionella effector functions during intracellular infection, we set out to phenotypically compare L. pneumophila and the largely unstudied pathogen L. dumoffii. L. dumoffii was found to share some of L. pneumophila's intracellular survival strategies, namely inhibition of the canonical endocytic pathway and creation of an ER-like intracellular niche. However, we identified two points of difference between the organisms: L. dumoffii's decoration with LC3 and lack of vacuole ubiquitination. Extending this insight, we show that LAP is the main process for LC3-decoration of L. dumoffii in host cells. LC3-decoration of LdCVs was also found to be Dot/Icm T4SS-dependent (Fig. 2). Thus, we propose that the majority of pathogenic (T4SS + ) L. dumoffii survive in host cells by creating an ER-like intracellular niche, but a portion of L. dumoffii are recognized by the host and decorated with LC3 in a ubiquitin-independent manner consistent with LAP (Fig. 10). Our data are consistent with the previous report that found ~20% of intracellular L. dumoffii were LAMP-1 positive at 4 hours post-infection in J774 mouse macrophages 84 .
The relationship between Dot/Icm-dependent LdCV maturation and LAP prompts us to question whether there is a link between phagosome maturation and LAP initiation. Regarding the balance of LAP-targeting over autophagy, our data indicates that vacuolar integrity is a likely contributing factor. Based on the lack of galectin-3 (or any other galectins we tested) association with LdCVs (Fig. 6), LAP-activation appears independent of membrane damage. Thus, vacuolar integrity of LdCVs may skews recognition away from membrane-damage responsive autophagy 25,79 . Alternatively, it remains possible that our membrane damage marker (galectin-3) is somehow removed from LdCVs. LAP's predominance may simply be due to the absence of phagosomal ubiquitin. In this scenario, TLR-and DAG-signaling and delayed phagosome maturation would provide positive LAP-activation signals and ubiquitin-negative intact phagosomes would avoid activation of classical autophagy.
Why LdCVs are subjected to LAP but LpCVs are not is an intriguing question. The L. pneumophila strain Lp01 used in this study produces the effector protein RavZ, which has an activity to irreversibly deconjugate LC3 and thus inhibit autophagy-related pathways including macroautophagy and LAP 17,78 . We examined the behavior of a RavZ deletion derivative and found that RavZ is not required for the LAP avoidance by L. pneumophila (Fig. 9A,B). This suggests that either L. pneumophila has additional effectors to inhibit LAP or L. dumoffii has an effector to promote LAP. The effector Spl 36 is one of such candidates which may inhibit LAP, however, L. dumoffii appears to encode the Spl ortholog. On the other hand, L. dumoffii ectopically producing RavZ failed to engage in LAP (Fig. 9C), indicating that L. dumoffii lacks sophistication to have a single effector protein like RavZ which efficiently blocks autophagy-related pathways, unlike some strains of L. pneumophila. Clarification of molecular mechanisms underlying L. dumoffii engagement in LAP awaits future study.
A remaining question is the physiological importance of LAP for bacterial restriction in the context of other immune defense mechanisms. Recently, LAP has been proposed to facilitate enhanced antigen presentation by delaying phagosome maturation 85 . It is tempting to speculate that in higher eukaryotes, LAP-mediated enhanced antigen presentation by macrophages infected with L. dumoffii engages innate or adaptive responses that limit disease progression, whereas the LAP-avoiding pathogen like L. pneumophila avoids this enhanced host response and causes more disease. Though the majority (~80%) will subvert canonical endocytic maturation and go on to replicate inside ER-like compartments, some (~20%) will be degraded following initiation of LC3-associated phagocytosis (LAP). Thus, the nature and kinetics of unproductive L. dumoffii trafficking differ substantially from those of L. pneumophila.
Rubicon is a Class III PI(3) kinase-associated protein 86 , which recruits PI(3) kinase to phagosomes and supports prolonged localization of PI3P 78 . Then Rubicon interacts with the NOX2 complex recruited in a PI3P-mediated manner and stabilizes the NOX2 complex, resulting induction of ROS production 78 . In contrast, LpCVs were reported to acquire PI3P right after infection in a model natural host Dictyostelium discoideum 87 , but soon PI4P becomes a dominant phosphoinositide species, which possibly involves the concerted actions of host PI(4) kinases, and host and bacterial phosphoinositide phosphatases [88][89][90][91] . Currently phosphoinositide dynamics on LdCVs are not known, and future studies focusing on PI3P pool on LAP-targeted versus LAP-escaping LdCVs will shed light on the requirement in LAP induced by fungal and bacterial infection, and the difference in fates of LdCVs and LpCVs. Furthermore LAP-targeting of LdCVs may be used as a model system examining the finer mechanistic details of LAP, particularly induced by bacterial infection.

Materials and Methods
Reagents. Unless otherwise noted, all chemicals were purchased from Sigma. Rapamycin was purchased from Santa Cruz (sc-3504). Restriction and molecular cloning enzymes were purchased from New England Biolabs or Toyobo Co. Ltd. Primary antibodies used include rabbit polyclonal antibodies to L. pneumophila Bacterial Strains. Bacterial strains used in this study are described in Table S1. Legionella strains were grown on charcoal-yeast extract plates or in aces-buffered yeast extract broth as already described 92 . When required, drugs were included in the bacteriological media at the following concentrations: for strains of Legionella species, streptomycin 100 μ g/ml, chloramphenicol 3 μ g/ml and kanamycin 10 μ g/ml; for strains of E. coli, ampicillin 100 μ g/ml, kanamycin 25 μ g/ml and chloramphenicol 20 μ g/ml; for Salmonella enterica serovar Typhimurium (S. Typhimurium), chloramphenicol 20 μ g/ml.

DNA manipulations.
Bacterial plasmids used in this study are listed in Table S2. pMMB207mCherry was created by cloning mCherry into pAM239. To induce constitutive expression of mCherry an internal region of the lacI repressor was disrupted. To create the ∆ dotA, ∆ flaA L. dumoffii and ∆ ravZ L. pneumophila mutants, the plasmids pSR47SLd∆ dotA, pSR47SLd∆ flaA and pSR47SLd∆ ravZ were created using primer pairs list in Table S3. Clones were sequenced using the primers pSR47Sac1 and pSR47SXba1. The ravZ C258A mutation was introduced using QuickChange II Site-Directed Mutagenesis Kit (Agilent Technologies) according to the manufacture's instruction. The lentiviral vector for the stable expression of GFP-LC3 was generated as follows. The genes encoding GFP and LC3 were amplified using the primer pairs listed in Table S3 with pEGFP-LC3 93 as a template. The PCR product was subsequently amplified with attB adaptor primers and cloned into pDONR201 (Invitrogen) with Gateway BP Clonase II Enzyme Mix (Invitrogen). The sequence encoding GFP-LC3 was then transferred to the lentiviral expression vector pLEXEF.pur 94 , using Gateway LR Clonase II Enzyme Mix (Invitrogen), to obtain pLEXEF.pur.GFP-LC3.
Genetic manipulations of bacteria. Strain Ld00 is a spontaneous streptomycin-resistant mutant of L. dumoffii NY23. Chromosomal deletion mutants of L. dumoffii were made using the pSR47S-based plasmids described above and following the method previously described for L. pneumophila 95 . Legionella strains constitutively expressing mCherry were created by introduction of pMMB207::mCherry by electroporation using the conditions described for L. pneumophila 2 . S. Typhimurium expressing pMMB207::mCherry was also created by electroporation and selection on chloramphenicol resistance.
Tissue culture and media. HEK293 cells stably expressing the Fcγ RII receptor were obtained from the laboratory of Craig Roy (Yale University, USA) 96 . HeLa cells lines stably expressing the Fcγ RII were created and kindly provided by Dr. Kohei Arasaki (Tokyo University of Pharmacy and Life Sciences, Japan). Wild-type MEF cells were provided by Dr. Miwa Sasai (Osaka University, Japan). Immortalized Atg7-deficient MEFs were kindly provided by Dr. Tatsuya Saitoh (Osaka University, Japan) on behalf of Dr. Masaaki Komatsu (Niigata University, Japan). TLR2 97 , TLR4 98 , Rubicon 99 KO mice were previously described. Murine BMDMs were obtained from mice as previously described 100 , and used fresh or stored as described 101 . All mammalian cell lines were maintained in 5% CO 2 at 37 °C. Mouse embryonic fibroblasts (MEF), HeLa, HEK293 and Fcγ RII-expressing stable variants 96 were maintained in Dulbecco's modified Eagle's growth medium (Gibco, Life Technologies) supplemented with 10% fetal calf serum. THP-1 cells were routinely passaged as non-adherent cells. RAW264.7 and THP-1 cells were cultured in RPMI 1640 medium (Gibco, Life Technologies) supplemented with 10% fetal calf serum. For knockdown experiments, HEK293 cells were seeded into 24-well tissue culture plates to achieve 40% confluency 24 hours later when plasmid DNA and siRNA was added to wells. Two-days later, media was changed to DMEM plus 5% FBS for the final overnight incubation. Infection experiments in complete media were performed at 72 hours post-transfection. For infection experiments with THP-1 cells, cells were differentiated in tissue culture wells at densities of 2 × 10 5 cells per well of the 24-well dishes and 5 × 10 4 cells for the 96-well dishes. Three days prior to use, THP-1 cells were incubated in media containing phorbol 12-myristate 13-acetate (PMA, Sigma). After 24-48 hours media was replaced with fresh media lacking PMA. The next day, infection experiments were performed. To create the RAW 264.7 cell line stably expressing GFP-LC3, the ViraPower Lentiviral Expression System (Invitrogen) was used to produce lentiviruses, and the resulting viral supernatant was used to transduce the lentiviral construct into RAW 264.7 cells. Subsequently, stable GFP-LC3-expressing cells were established by selection with puromycin (Clontech). RAW cells stably expressing GFP-LC3 were cultured in RPMI 1640 medium supplemented with 10% fetal calf serum.
Host cell manipulations. HEK293 and HeLa cell transfections were performed using Lipofectamine 2000 (Invitrogen) or Lipofectamine LTX+ (Invitrogen) according to the manufacturer's instructions. For co-transfection assays in HEK293 cells, plasmid DNA (200 ng) was mixed with siRNA (50 nM) and Lipofectamine 2000 before co-transfection into cells. Scrambled siRNA and Ulk1 siRNA for RNAi experiments were ordered from Sigma Genosys. The sequence of the Ulk1 siRNA used was previously reported 102 . Lipofectamine LTX+ (Invitrogen) was used to transfect MEF cells using 1.75 μ l of LTX, 0.5 μ l 'plus' reagent and 0.5 μ g DNA per well of the 24-well dishes.
Bacterial infections. Infections with L. pneumophila or L. dumoffii were performed with bacteria growth in 2-ml AYE both cultures for 20 hours. The final OD 600 of L. dumoffii cultures used for infection was ~2.8 (range of 2.7 to 3.2), whereas L. pneumophila cultures were used at ~4.0 (range of 3.8 to 4.4). HEK293 Fcγ RII and HeLa Fcγ RII cells were infected with L. dumoffii at a multiplicity of infection (moi) of one as previously described for L. pneumophila 90 . For non-Fcγ RII-mediated uptake, an moi of 100 to 300 was used. For infection of RAW264.7 cells stably expressing GFP-LC3, THP-1 and MEF cells an moi of 10 to 30 was used. For dual infection experiments of L. pneumophila mCherry and L. dumoffii, 10 times more L. pneumophila was used than L. dumoffii because we observed better uptake of L. dumoffii than L. pneumophila. To propagate Salmonella for infections, overnight cultures were diluted to OD 600 of 0.17 and grown for ~3 hours until the OD 600 reached 1.8-2.5, and finally bacteria were added to cells at an moi of 100. At 20 min post-infection, 100 μ g/ml gentamycin was added and maintained until cells were washed and fixed. In all experiments cells were fixed with 4% PFA for 15 min at 37 °C. After staining coverslips were stained with Hoechst at 1 μ g/ml and mounted using the ProLong gold antifade reagent (Invitrogen). For growth curves in mammalian cells an MOI of 0.5 was used and performed as previously described 90 . For experiments using C57BL/6 BMDM and MEF cells, infections were performed using L. dumoffii∆ flaA.
RT-PCR. Knockdown efficiency was validated by quantitative RT-PCR using Thunderbird TM SYBR ® qPCR mix on cDNA synthesized by the SuperScript TM III (Invitrogen#18080-051) kit from total RNA extracted from cells at 72 h post-knockdown using the RNeasy kit (Qiagen). The primer pairs used to assess transcript levels are listed in Table S2.
Microscopy. Epifluorescence micrographs were taken using a TE2000 (Nikon) inverted microscope and this microscope was used for all counting experiments. Confocal micrographs were taken using a LSM510 microscope (Zeiss) with a 100x/1.4 numerical aperture objective, or a Fluoview FV10i (Olympus) microscope with a 60× /1.35 numerical aperture objective. Correlative light electron microscopy (CLEM) experiments were performed as previously described 103 . Basically, RAW264.7 cells stably expressing EYGP-LC3 were cultured on glass bottom dishes (with grids) and then infected with mCherry-expressing Legionella dumoffii for 2.5-3 hours. After fixing, confocal images were taken before processing the samples for electron microscopy.

Survival assays in MEFs.
To perform the L. dumoffii survial assay in Atg7+ and Atg7-MEF cells, MEFs were seeded at 5 × 10 4 per well in 24-well format -to reach ~80% confluency at the time of infection. Fresh 2-day heavy patch L. dumoffii∆flaA expressing mCherry were grown overnight in AYE plus chloramphenicol broth until OD 600 reached 2.8-3.0. Bacteria were added to wells at 2 × 10 5 per well. Plates were spun at 1000 rpm for 5 minutes and then incubated for 5 hours under standard cell culture conditions. At 5 hours post-infection, wells were gently washed two times with PBS to remove extracellular bacteria. Cells were then lysed in 1 ml of sterile water per well. To promote host cells lysis, cells were freezed and thawed before vigorously resuspending the contents of each well and plating for bacterial viability on CYE media. For each cell type, sixteen independent wells were assessed from four separate plates and each well was plated in duplicate. Data are shown compared to control cells (wild-type) average bacterial survival, which was normalized to 100%. Significance was determined by the students' T-Test.
Survival Assays in Macrophages. Differentiated BMDMs were seeded in 24-well plates at 1.5 × 10 5 cells per well. The following day, L. dumoffii∆flaA expressing mCherry were added at 5 × 10 5 per well and spun for 5 min at 1000 rpm. At 5 hours post-infection, wells were treated as described for MEF cells with a final well volume of 1 ml. One hundred μ l of 1/100 dilutions were plated to enumerate bacterial survival. In triplicate experiments, six wells were assessed for each cell type. Data are shown compared to control cells (wild-type) average bacterial survival, which was normalized to 100%. Significance was determined by the students' T-Test.
Statistical Analysis. To statistically assess significance, calculations were performed using the paired Student's t-test (homoscedastic two-tailed, paired) using Excel software (Microsoft). In all graphs error bars represent standard error of the mean (SEM). Where appropriate, p values or ns for not significant were denoted.
Ethics statement. All animal experiments were performed in accordance with the institutional guidelines and were approved by the Animal Care and Use Committee of the Research Institute for Microbial Diseases, Osaka University, Japan (Biken-AP-H26-10-0, IFReC-AP-H27-07-0).