ER remodeling by the large GTPase atlastin promotes vacuolar growth of Legionella pneumophila

Abstract

The pathogenic bacterium Legionella pneumophila replicates in host cells within a distinct ER‐associated compartment termed the Legionella‐containing vacuole (LCV). How the dynamic ER network contributes to pathogen proliferation within the nascent LCV remains elusive. A proteomic analysis of purified LCVs identified the ER tubule‐resident large GTPase atlastin3 (Atl3, yeast Sey1p) and the reticulon protein Rtn4 as conserved LCV host components. Here, we report that Sey1/Atl3 and Rtn4 localize to early LCVs and are critical for pathogen vacuole formation. Sey1 overproduction promotes intracellular growth of L. pneumophila, whereas a catalytically inactive, dominant‐negative GTPase mutant protein, or Atl3 depletion, restricts pathogen replication and impairs LCV maturation. Sey1 is not required for initial recruitment of ER to PtdIns(4)P‐positive LCVs but for subsequent pathogen vacuole expansion. GTP (but not GDP) catalyzes the Sey1‐dependent aggregation of purified, ER‐positive LCVs in vitro. Thus, Sey1/Atl3‐dependent ER remodeling contributes to LCV maturation and intracellular replication of L. pneumophila.

Introduction

The ubiquitous environmental bacterium Legionella pneumophila is the causative agent of the severe pneumonia Legionnaires’ disease 1, 2. Legionella pneumophila is an opportunistic pathogen that employs a largely conserved mechanism to replicate in environmental protozoa as well as in macrophages within a unique endoplasmic reticulum (ER)‐associated compartment termed the Legionella‐containing vacuole (LCV) 3, 4, 5, 6. LCV formation requires the L. pneumophila Icm/Dot (intracellular multiplication/defective organelle trafficking) type IV secretion system (T4SS). This T4SS translocates the astonishing number of 300 different so‐called effector proteins into host cells, where they subvert signal transduction and vesicle trafficking pathways 4, 7, 8, 9, 10, 11.

Legionella‐containing vacuoles avoid fusion with bactericidal lysosomes, but extensively communicate with the endosomal and secretory vesicle trafficking pathways, eventually associating with the ER in an intimate manner 12, 13, 14, 15, 16. After phagosome closure, the LCV acquires ER membrane markers such as HDEL‐/KDEL‐GFP or calnexin‐GFP 15, 17, 18. The pathogen vacuole then moves along microtubules and undergoes a transition from a “tight” to a “spacious” compartment, where the vacuole membrane detaches from the bacteria and stays connected only at the bacterial poles, if at all. LCVs intercept early secretory vesicles at ER exit sites, a process that requires the activity of the small GTPases Sar1, Arf1, and Rab1 19, 20, 21. ER‐derived vesicles fuse with LCVs by non‐canonical pairing of the ER v‐SNARE Sec22b with plasma membrane‐derived t‐SNAREs (syntaxins, SNAP23) on the pathogen vacuole 22. Moreover, LCVs are decorated with several other small GTPases of the Rab family, Rap1 and Ran, many of which are implicated in intracellular replication of L. pneumophila 23, 24, 25, 26.

In the genetically tractable social soil amoeba Dictyostelium discoideum, LCVs seem not to fuse with the ER for a prolonged time, since even at 14 h post‐infection (p.i.), the luminal ER marker calreticulin‐GFP remains confined at the pathogen vacuole boundary and does not diffuse into its lumen 15. Furthermore, LCVs undergo a PI conversion from PtdIns(3)P to PtdIns(4)P within 2 h p.i., and for at least 8 h p.i., the replication‐permissive, PtdIns(4)P‐positive LCV remains spatially separated from calnexin‐positive ER 27. In macrophages, LCVs also attach to (ribosome‐studded, rough) ER, which was reported to readily fuse with the pathogen compartment 16. In any case, the fusion of LCVs with the ER during the formation of a replication‐permissive vacuole appears dispensable, as L. pneumophila already grows within vacuoles that are attached to but not merged with the ER 18, 27. Taken together, LCV formation can be described as tri‐phasic process comprising the avoidance of lysosome fusion, interaction with early secretory vesicles, and attachment to ER.

The ER forms a complex and dynamic network of perinuclear, “rough” sheets and peripheral, “smooth” tubules 28, 29. Recent morphological and dynamic analysis using super‐resolution imaging revealed that the ER consists almost exclusively of tubules and structures termed ER “matrices” (formerly referred to as “sheets”) 30. The architecture of this network is maintained by the microtubule cytoskeleton, as well as by sheet‐ and tubule‐localizing resident ER proteins 31, 32. The ER sheet structure is jointly maintained by Rtn4 (reticulon 4) and Climp63 (cytoskeleton‐linking membrane protein of 63 kDa), while the tubule structure requires Rtn4 (Nogo4) and its interactor DP1/Yop1 33, 34.

To generate the ER tubular network, reticulon proteins interact with dynamin‐like large GTPases of the atlastin family 35, which is conserved and called Sey1p (Synthetic enhancement of Yop1) in S. cerevisiae and plants 36, 37. Atlastin/Sey1 proteins consist of a large N‐terminal guanosine triphosphatase (GTPase) domain, followed by a three‐helix bundle (3HB), two adjacent transmembrane motifs (TMs), and a cytosolic C‐terminal domain (CT). The consensus sequence of the active site phosphate‐binding loop (P‐loop) of these large GTPases includes the conserved GxxxxGKS motif (Fig 1A). Mammals produce three isoforms of atlastin (Atl1‐3) that show tissue‐specific distribution: While Atl1 is produced preferentially in neuronal tissue, Atl2 and Atl3 are ubiquitously produced 38. Atlastins are intrinsic membrane proteins that dimerize in trans (different membranes), thus catalyzing homotypic membrane fusions and promoting the dynamic remodeling of the ER network 39. Here, we assess the role of Sey1/Atl3 for LCV formation and intracellular replication of L. pneumophila.

Figure 1. The Dictyostelium discoideum atlastin3 homolog Sey1 localizes to LCVs.

1. Domain architecture of atlastin/Sey1 proteins. Atlastins consists of a large N‐terminal guanosine triphosphatase (GTPase) domain, followed by a three‐helix bundle (3HB), two adjacent transmembrane motifs (TMs), and a cytosolic C‐terminal domain (CT). The consensus sequence of the active site phosphate‐binding loop (P‐loop) of atlastin GTPases includes the conserved GxxxxGKS motif.
2. Dictyostelium discoideum Ax3 ectopically producing GFP‐Sey1 was infected (MOI 10, 1 h) with mCerulean‐producing L. pneumophila JR32 or ΔicmT (pNP99), fixed with PFA and labeled with anti‐calnexin (Caln) and anti‐SidC antibodies; scale bars: 5 μm (main image), 1 μm (insert).
3. Quantification of GFP‐Sey1‐positive LCVs in D. discoideum at 1 h post‐infection (p.i.); 100 infected cells per sample were counted each in three independent experiments (mean and standard error of mean, SEM; ***P < 0.001, Student's t‐test).
4. The GTPase activity of the purified GTPase domains (5 μM) of GST‐Sey1p (S. cerevisiae), GST‐Sey1 or GST‐Sey1_K154A (D. discoideum) is shown as μM phosphate released over time. Heat‐inactivated (h.i.) protein was used as a control. Data are representative of two technical replicates each of two separate protein preparations.

Results

Sey1 localizes to LCVs and modulates replication of Legionella pneumophila in Dictyostelium discoideum

In a previous study, we identified the large dynamin‐like GTPases Sey1 and Atl3 in the proteome of LCVs purified from infected D. discoideum (DDB_G0279823) or RAW 267.4 macrophages (NP_001156977) 24, respectively. The D. discoideum gene DDB_G0279823 is annotated as “Sey1”, but has not been characterized thus far. The corresponding protein (Q54W90) shares a domain architecture identical to S. cerevisiae Sey1p and mammalian atlastins, comprising the GTPase, 3HB, TM and CT domains (Fig 1A). Thus, D. discoideum Sey1 likely adopts similar functions as the yeast and mammalian counterparts. In D. discoideum, GFP‐Sey1 colocalizes with calnexin, but not with the PI lipid PtdIns(4)P (Figs 1B and EV1). To test whether Sey1 accumulates on LCVs, D. discoideum ectopically producing GFP‐Sey1 was infected with mCerulean‐producing L. pneumophila JR32 or ΔicmT and immuno‐stained for calnexin and SidC, an Icm/Dot‐translocated effector decorating the LCV membrane (Fig 1B). Quantification of GFP‐Sey1‐positive LCVs in D. discoideum at 1 h p.i. revealed that close to 90% of pathogen vacuoles harboring the parental strain, but only about 10% of vacuoles harboring the ΔicmT mutant strain, were decorated with the large GTPase (Fig 1C).

Figure EV1. Localization of GFP‐Sey1 or GFP‐Sey1_K154A and ER architecture.

• A–D
  (A, C) Confocal fluorescence microscopy of D. discoideum Ax3 producing GFP‐Sey1 or GFP‐Sey1_K154A, fixed with PFA, and immuno‐labeled with an (A) anti‐calnexin (Caln) antibody, or (C) anti‐PtdIns(4)P antibody, scale bars: 10 μm. (B, D) Determination of Pearson's correlation coefficient of GFP‐Sey1 or GFP‐Sey1_K154A versus (B) Caln or (D) PtdIns(4)P using Coloc 2 from Fiji (ImageJ). Data show individual data points of one experiment (n = 50) and are representative of two independent experiments.

Dictyostelium discoideum Sey1 harbors a conserved lysine residue at position 154 in the predicted nucleotide‐binding P‐loop of the GTPase domain. Mutation of the P‐loop lysine to an alanine residue is expected to yield a catalytically inactive, dominant‐negative form that is impaired for GTP binding, as shown for S. cerevisiae Sey1p 35 (Appendix Fig S1). Upon production in D. discoideum, GFP‐Sey1_K154A did not colocalize with calnexin or PtdIns(4)P (Fig EV1). To test whether D. discoideum Sey1 exhibits GTPase activity and K154 is required, we produced in E. coli and purified N‐terminal fragments of Sey1 and Sey1_K154A (aa 1‐630) corresponding to the GTPase domain of Sey1p (aa 1‐498) 36. Indeed, Sey1_1‐630 showed GTPase activity (phosphate release) similar to Sey1p_1‐498, while Sey1_K154A_1‐630 or heat‐inactivated Sey1_1‐630 was inactive (Fig 1D). As expected, the release of phosphate by Sey1_1‐630 was dependent on GTP and Mg²⁺ (data not shown). Taken together, D. discoideum Sey1 shares the domain architecture and GTPase activity with atlastins and localizes to the ER as well as to LCVs. Therefore, Sey1 represents the bona fide D. discoideum atlastin.

To further assess the role of D. discoideum Sey1 for L. pneumophila infection, amoebae producing GFP‐Sey1 or GFP‐Sey1_K154A were infected with DsRed‐producing L. pneumophila JR32 parental or mutant strains, and localization of the fusion proteins was imaged by confocal fluorescence microscopy (Figs 2A and EV2). While GFP‐Sey1 accumulated on LCVs harboring strain JR32 as early as 1 h and for at least 8 h, the GFP‐Sey1_K154A mutant protein barely localized to LCVs during the entire infection. At 1 h p.i., more than 90% of LCVs harboring strain JR32 were decorated with GFP‐Sey1; yet, less than 20% of pathogen vacuoles harboring JR32 or ΔicmT accumulated GFP‐Sey1_K154A (Figs 2B and EV2). At 8 h p.i., vacuoles harboring the mutant strains ΔicmT or ΔsidC‐sdcA (defective for ER accumulation) were neither decorated with GFP‐Sey1 nor with GFP‐Sey1_K154A (Fig EV2).

Figure 2. Sey1 modulates replication of Legionella pneumophila in Dictyostelium discoideum .

• A
  Dictyostelium discoideum Ax3 producing GFP‐Sey1 or GFP‐Sey1_K154A was infected (MOI 10) for the time indicated with mCherry‐producing L. pneumophila JR32 (pNP102), fixed with PFA, stained with anti‐calnexin (Caln) antibody and DAPI, and imaged by confocal fluorescence microscopy; scale bars: 5 μm.
• B
  Quantification of colocalization in D. discoideum Ax3 producing GFP‐Sey1 or GFP‐Sey1_K154A, infected (MOI 10, 1‐8 h) with mCherry‐producing L. pneumophila JR32 or ΔicmT (pNP102) as in (A). In two independent experiments, > 200 pathogen vacuoles were scored at the time points indicated; data points represent means.
• C
  Dictyostelium discoideum Ax3 producing calnexin (CnxA)‐mCherry in the absence or presence of GFP‐Sey1 or GFP‐Sey1_K154A was infected (MOI 20, 2 h) with mCerulean‐producing L. pneumophila JR32 (pNP99), chemically fixed, and analyzed by transmission electron microscopy (arrows, limiting LCV membrane; arrowheads, ER double membrane attached to LCV membrane; scale bars: 2 μm (main image), 0.5 μm (insert)).
• D, E
  Dictyostelium discoideum Ax3 producing GFP, CnxA‐GFP, GFP‐Sey1, or GFP‐Sey1_K154A was infected (MOI 10, 24 h) with mCherry‐producing L. pneumophila JR32, ΔicmT, or ΔsidC‐sdcA (pNP102), and intracellular bacterial replication was assessed by colony‐forming units (CFU; relative intracellular replication). Mean and SEM of three independent experiments are shown (statistics refer to JR32‐infected D. discoideum strains; *P < 0.05, **P < 0.01, ***P < 0.001, Student's t‐test).

Figure EV2. LCVs harboring Legionella pneumophila ΔicmT or ΔsidC‐sdcA do not acquire Sey1.

• A–C
  Dictyostelium discoideum Ax3 producing GFP‐Sey1 or GFP‐Sey1_K154A was infected (MOI 10, 2 h or 8 h) with mCherry‐producing L. pneumophila (A) JR32, (B) ΔicmT, or (C) ΔsidC‐sdcA (pNP102), fixed with PFA, stained with an anti‐calnexin (Caln) antibody and DAPI, and imaged by confocal fluorescence microscopy; scale bars: 5 μm.

We also analyzed the effect of GFP‐Sey1 or GFP‐Sey1_K154A on the morphology and architecture of LCVs by electron microscopy (EM). To this end, D. discoideum producing calnexin‐mCherry in the absence or presence of GFP‐Sey1 or GFP‐Sey1_K154A was infected with L. pneumophila JR32, chemically fixed and visualized by EM. This high‐resolution analysis confirmed that dependent on endogenous Sey1 or upon overproduction of GFP‐Sey1, the ER tightly associates with the limiting LCV membrane (Fig 2C). In contrast, overproduction of GFP‐Sey1_K154A abolished the accumulation of ER around the pathogen vacuole.

Next, we tested whether the overproduction of Sey1 or Sey1_K154A in D. discoideum affects intracellular growth of L. pneumophila. To this end, D. discoideum amoebae producing GFP, calnexin (CnxA)‐GFP, GFP‐Sey1, or GFP‐Sey1_K154A were infected with mCherry‐producing L. pneumophila JR32. Intracellular bacterial replication was assessed by colony‐forming units (CFU; Fig 2D) or fluorescence (Appendix Fig S2). CFU as well as fluorescence measurements indicated that the overproduction of Sey1 significantly promotes the intracellular replication of L. pneumophila, while the overproduction of catalytically inactive (dominant‐negative) Sey1_K154A substantially reduced intracellular growth. In contrast, neither Sey1 nor Sey1_K154A affected the uptake of L. pneumophila JR32 or ΔicmT by D. discoideum (Appendix Fig S2).

We also assessed whether Sey1 or Sey1_K154A affects the intracellular replication of L. pneumophila lacking the Icm/Dot‐translocated effector SidC and its paralog SdcA. The ΔsidC‐sdcA mutant strain is not impaired for intracellular replication; yet, LCVs harboring the mutant exhibit slower ER acquisition kinetics 18, as well as less extensive ER contact areas 27, and they barely accumulate GFP‐Sey1 or GFP‐Sey1_K154A (Fig EV2). Interestingly, while dominant‐negative Sey1_K154A restricts the intracellular replication of the ΔsidC‐sdcA mutant, the overproduction of Sey1 does not enhance intracellular growth of the mutant in contrast to the parental strain JR32 (Fig 2E). These results suggest that the tight association of the ER to the LCV catalyzed by SidC/SdcA is required for Sey1 to exert a growth‐promoting effect.

Taken together, the D. discoideum Atl3 homologue, Sey1 (but not catalytically inactive Sey1_K154A), is recruited to pathogen vacuoles harboring L. pneumophila JR32 but not the ΔicmT or ΔsidC‐sdcA mutant strain. Furthermore, the overproduction of Sey1 promotes intracellular replication of L. pneumophila JR32, but not ΔsidC‐sdcA, while a catalytically inactive, likely dominant‐negative mutant protein reduces intracellular bacterial growth.

Recruitment of ER to PtdIns(4)P‐positive LCVs in dually labeled Dictyostelium discoideum

In order to dissect the specific step where Sey1 plays a role in LCV maturation, we employed dually labeled D. discoideum strains. To assess the localization of the large GTPase to PtdIns(4)P‐positive LCVs containing L. pneumophila JR32, we used D. discoideum producing P4C‐mCherry and either GFP‐Sey1 or GFP‐Sey1_K154A (Fig 3A). Strikingly, live‐cell confocal fluorescence microscopy revealed that GFP‐Sey1 surrounded PtdIns(4)P‐positive LCVs in a punctate manner, such that the limiting LCV membrane (defined by PtdIns(4)P) and the ER membrane were clearly distinct. While GFP‐Sey1_K154A localized throughout the cell in a diffuse pattern, but not to the LCV, the labeling of LCVs with P4C‐mCherry appeared at least as intense in D. discoideum producing GFP‐Sey1_K154A compared to amoebae producing GFP‐Sey1. Together, these results indicate that the PI conversion and decoration with PtdIns(4)P of LCVs proceeds independently of Sey1 activity.

Figure 3. Recruitment of ER to PtdIns(4)P‐positive LCVs in dually labeled Dictyostelium discoideum .

1. Dictyostelium discoideum Ax3 producing P4C‐mCherry and GFP‐Sey1 or GFP‐Sey1_K154A was infected (MOI 10, 2 h) with mCerulean‐producing L. pneumophila JR32 (pNP99) and imaged by live‐cell confocal fluorescence microscopy; scale bars: 3 μm.
2. Imaging flow cytometry (IFC) of D. discoideum Ax3 producing P4C‐mCherry and GFP‐Sey1 or GFP‐Sey1_K154A, infected (MOI 5) for the time indicated with mPlum‐producing L. pneumophila JR32, ΔicmT, or ΔsidC‐sdcA (pAW14). Mean IFC colocalization score for P4C‐mCherry for > 700 LCVs is depicted in one data point. The data are representative of two independent experiments.

These results were quantified by imaging flow cytometry (IFC) using dually labeled D. discoideum strains producing P4C‐mCherry and either GFP‐Sey1 or GFP‐Sey1_K154A, infected with L. pneumophila JR32, ΔicmT, or ΔsidC‐sdcA mutant bacteria producing mPlum (Fig EV3). The IFC colocalization score (see Materials and Methods) for the acquisition of P4C‐mCherry to LCVs harboring L. pneumophila JR32 or ΔsidC‐sdcA was similar (for the mutant strain slightly higher) and increased over time independently of Sey1 function (Fig 3B). The P4C‐mCherry intensity was even higher in the GFP‐Sey1_K154A background, similar to what was observed by confocal fluorescence microscopy (Fig 3A). Hence, the accumulation of PtdIns(4)P on LCVs neither requires host Sey1 nor the bacterial effectors SidC/SdcA. The IFC colocalization score was lower and did not increase over time for LCVs harboring ΔicmT mutant bacteria, confirming that these LCVs do not undergo PI conversion and are not decorated with PtdIns(4)P 27. Taken together, P4C‐mCherry accumulates at the limiting membrane of LCVs in D. discoideum regardless of whether the amoebae produce GFP‐Sey1 or GFP‐Sey1_K154A. Therefore, the initial early formation of PtdIns(4)P‐positive LCVs is independent of the large GTPase.

Figure EV3. Imaging flow cytometry of LCV host factor acquisition in Dictyostelium discoideum .

• A, B
  Imaging flow cytometry of D. discoideum Ax3 producing GFP‐Sey1 or GFP‐Sey1_K154A and (A) P4C‐mCherry or (B) calnexin (CnxA)‐mCherry, infected (MOI 5, 4 h) with mPlum‐producing L. pneumophila JR32, ΔicmT, or ΔsidC‐sdcA (pAW14). 10,000 events were analyzed, and colocalization of GFP (Sey1/Sey1_K154A) or mCherry (P4C/CnxA) with mPlum (L. pneumophila, Lpn) is shown; scale bars: 7 μm.

Analogously, to assess the colocalization of Sey1 with calnexin‐positive LCVs, we used dually labeled D. discoideum strains producing calnexin‐mCherry and either GFP‐Sey1 or GFP‐Sey1_K154A. The amoebae were infected with mCerulean‐producing L. pneumophila JR32 and analyzed by live‐cell confocal fluorescence microscopy (Fig 4A). In the tandem labeled amoebae, GFP‐Sey1 colocalized with calnexin‐mCherry, which surrounds the PtdIns(4)P‐positive LCV membrane. In contrast, GFP‐Sey1_K154A localized throughout the cell in a more diffuse pattern and did not colocalize with calnexin‐mCherry around the pathogen vacuole.

Figure 4. Recruitment of calnexin‐positive ER to LCVs in dually labeled Dictyostelium discoideum .

• A
  Dictyostelium discoideum Ax3 producing calnexin (CnxA)‐mCherry and GFP‐Sey1 or GFP‐Sey1_K154A was infected (MOI 10, 2 h) with mCerulean‐producing L. pneumophila JR32 (pNP99) and imaged by live‐cell confocal fluorescence microscopy; scale bars: 3 μm.
• B, C
  Imaging flow cytometry (IFC) of D. discoideum Ax3 producing CnxA‐mCherry and GFP‐Sey1 or GFP‐Sey1_K154A, infected (MOI 5) for the time indicated with mPlum‐producing L. pneumophila JR32, ΔicmT, or ΔsidC‐sdcA (pAW14). Mean IFC colocalization score for (B) calnexin‐mCherry or (C) GFP‐Sey1/GFP‐Sey1_K154A for > 600 LCVs is depicted in one data point. The data are representative of at least two independent experiments.

To quantify the microscopy results, we infected dually labeled D. discoideum strains producing calnexin‐mCherry and either GFP‐Sey1 or GFP‐Sey1_K154A with L. pneumophila JR32, ΔicmT, or ΔsidC‐sdcA mutant bacteria producing mPlum and assessed 10,000 cells each at three different time points using IFC (Fig EV3). The IFC colocalization score for the acquisition of calnexin‐mCherry to LCVs harboring L. pneumophila JR32 increased over time and was very similar in amoebae producing GFP‐Sey1 or GFP‐Sey1_K154A (Fig 4B). These results indicate that calnexin‐mCherry accumulates at the cytoplasmic side of LCVs regardless of Sey1 function, and thus, Sey1 does not seem to affect the initial recruitment of ER to LCVs. In amoebae producing GFP‐Sey1 or GFP‐Sey1_K154A, the calnexin‐mCherry IFC colocalization score was lower for LCVs harboring ΔsidC‐sdcA or ΔicmT mutant bacteria and did not increase over time for the latter. Thus, the mutant strains acquire less (or no) calnexin, again independently of Sey1. Moreover, LCVs harboring L. pneumophila JR32 or ΔsidC‐sdcA (but not ΔicmT) mutant bacteria acquired GFP‐Sey1, but only JR32 LCVs stained (faintly) for GFP‐Sey1_K154A (Fig 4C). In summary, the recruitment of calnexin and Sey1 to LCVs appears to proceed independently of Sey1 activity.

Sey1 promotes expansion of PtdIns(4)P‐positive LCVs

A potential role of Sey1 for expansion of LCVs was first assessed by infecting D. discoideum producing CnxA‐GFP, GFP‐Sey1, or GFP‐Sey1_K154A with mCherry‐producing L. pneumophila JR32 (Appendix Fig S3). The infected amoebae were homogenized and fixed, and LCV expansion was analyzed by confocal fluorescence microscopy. In the course of 8 h, CnxA‐GFP or GFP‐Sey1 but not GFP‐Sey1_K154A accumulated on LCVs and the diameter of the pathogen vacuole increased.

To further validate these findings, and to determine whether the activity of Sey1 influences the expansion of individual PtdIns(4)P‐positive LCVs, we used dually labeled D. discoideum producing CnxA‐GFP (Fig 5A), GFP‐Sey1 (Fig 5B), or GFP‐Sey1_K154A (Fig 5C) in parallel with P4C‐mCherry. The amoebae were infected with mCerulean‐producing L. pneumophila JR32, and the diameter of PtdIns(4)P‐positive LCVs was measured in homogenates at early (1 h) and late time points (8 h) p.i. The diameter of individual PtdIns(4)P‐positive LCVs increased over time in D. discoideum producing CnxA‐GFP and P4C‐mCherry (Fig 5A and D). Upon overproduction of GFP‐Sey1, the mean diameter of PtdIns(4)P‐positive LCVs was significantly larger compared to D. discoideum producing CnxA‐GFP (Fig 5B and D). Strikingly, in D. discoideum producing catalytically inactive (dominant‐negative) GFP‐Sey1_K154A, the mean diameter of PtdIns(4)P‐positive LCVs did not increase over time (Fig 5C and D). These results are in agreement with a role for Sey1 in the expansion of PtdIns(4)P‐positive LCVs, thereby promoting LCV maturation and intracellular replication (Fig 2C).

Figure 5. Sey1 promotes expansion of PtdIns(4)P‐positive LCVs.

• A–C
  Dictyostelium discoideum Ax3 producing (A) calnexin (CnxA)‐GFP, (B) GFP‐Sey1, or (C) GFP‐Sey1_K154A and P4C‐mCherry was infected (MOI 10) for the time indicated with mCerulean‐producing L. pneumophila JR32 (pNP99), lysed, fixed with PFA, and imaged by confocal fluorescence microscopy; scale bars: 1 μm.
• D
  Quantification of the diameter of 50 individual PtdIns(4)P‐positive LCVs from the experiment described in (A–C); means and SEM of three independent experiments are shown (*P < 0.05, ***P < 0.001, Student's t‐test).

GTP‐ and Sey1‐dependent LCV aggregation in vitro

Since the initial recruitment of ER to PtdIns(4)P‐positive LCVs also occurs in the presence of catalytically inactive, presumably dominant‐negative Sey1_K145A, we sought to establish a more reductionist system to analyze the steps of LCV maturation that do require Sey1. To this end, we purified intact LCVs from infected D. discoideum employing the well‐established two‐step protocol comprising immuno‐magnetic separation and Histodenz density gradient centrifugation 23, 24, 26, 40. The protocol was applied to purify intact LCVs harboring mCherry‐producing L. pneumophila JR32 from D. discoideum producing calnexin‐GFP, GFP‐Sey1, or GFP‐Sey1_K154A (Fig 6A). While the majority of all LCVs were decorated with SidC, GFP‐Sey1, or calnexin‐GFP, GFP‐Sey1_K154A did not decorate the purified pathogen vacuoles.

Figure 6. GTP‐ and Sey1‐dependent LCV aggregation in vitro .

• A
  Close‐up fluorescence micrographs of purified LCVs from D. discoideum Ax3 producing calnexin (CnxA)‐GFP, GFP‐Sey1 or GFP‐Sey1_K154A, infected (MOI 10, 1 h) with mCherry‐producing L. pneumophila JR32 (pNP102). The LCVs were purified by immuno‐magnetic separation and Histodenz density gradient centrifugation (fractions 3–5) and stained with an anti‐SidC antibody; scale bars: 1 μm.
• B
  Fluorescence micrographs (and magnifications) of purified LCVs (density gradient fractions 3–5) without additive (control), or after addition of 5 mM GDP, GTP, or Gpp(NH)p (30 min, 30°C) in the presence of 4 mM MgCl₂; scale bars: 10 μm.
• C, D
  After addition of 5 mM GTP to purified LCVs, (C) the total number of aggregated LCVs (mCherry‐positive bacteria) per view field or (D) the diameter of LCV aggregates (CnxA‐GFP or GFP‐Sey1) per view field was quantified. Data represent means and SEM of three independent experiments (***P < 0.001, Student's t‐test; C: GFP‐Sey1_K154A compared to CnxA‐GFP).

To test biochemical properties, we added 5 mM of GDP, GTP, or the non‐hydrolysable analog Gpp(NH)p to these LCV preparations (Histodenz fractions 3–5) and tested the aggregation of LCVs by fluorescence microscopy. To this end, input fractions with almost identical yield were used (Fig EV4). Treatment with GTP, but not with GDP or Gpp(NH)p, caused a massive aggregation of the purified LCVs from D. discoideum producing GFP‐Sey1 or calnexin‐GFP, but not from amoebae producing GFP‐Sey1_K154A (Fig 6B). Moreover, while the number of aggregated LCVs from D. discoideum producing GFP‐Sey1 or calnexin‐GFP per view field was similar (Fig 6C), the average aggregate diameter was almost twofold larger for LCVs from D. discoideum producing GFP‐Sey1 (Figs 6D and EV4). Taken together, these results indicate that GTP (but not GDP or Gpp(NH)p) and Sey1 catalyze ER aggregation and remodeling around LCVs, thus contributing to LCV maturation.

Figure EV4. GTP‐ and Sey1‐dependent LCV aggregation in vitro .

1. Quantification of mCherry‐producing L. pneumophila JR32 (pNP102) in purified LCV preparations from D. discoideum producing calnexin (CnxA)‐GFP, GFP‐Sey1, or GFP‐Sey1_K154A. Data show mean and SEM of three independent experiments with 10 view fields each counted.
2. Representative examples of LCV aggregation. LCVs were purified from D. discoideum Ax3 producing GFP‐Sey1 or CnxA‐GFP, infected (MOI 10, 1 h) with L. pneumophila JR32, and treated with 5 mM GTP (30 min, 30°C) in the presence of 4 mM MgCl₂, scale bars: 10 μm.

Atl3 localizes to LCVs and promotes growth of Legionella pneumophila in macrophages

Previously, we identified Atl3 in the proteome of LCVs purified from infected murine RAW 267.4 macrophages (NP_001156977) 24. To validate the presence of Atl3 on LCVs in these macrophages, the cells were infected with DsRed‐producing L. pneumophila JR32 or avirulent ΔicmT strain and immuno‐labeled with anti‐Atl3 and anti‐Calr (calreticulin) antibodies (Fig 7A). The quantification of Atl3‐positive LCVs in the macrophages at 1 h p.i. revealed that close to 85% of the pathogen vacuoles harboring the parental strain, but only about 15% of the vacuoles harboring the ΔicmT mutant strain acquired the large GTPase (Fig 7B).

Figure 7. Atl3 localizes to LCVs and promotes growth of Legionella pneumophila in macrophages.

1. RAW 264.7 macrophages were infected (MOI 10, 1 h) with DsRed‐producing L. pneumophila JR32 or ΔicmT (pSW001), labeled with anti‐Atl3 (atlastin3) and anti‐Calr (calreticulin) antibodies and DAPI; scale bars: 10 μm (main image), 1 μm (inset).
2. Quantification of Atl3‐positive LCVs in RAW 264.7 macrophages at 1 h p.i. Mean and SEM of three independent experiments are shown (n = 50; **P < 0.01, Student's t‐test).
3. RAW 264.7 macrophages transfected for 24 h with 150 nM siRNA oligonucleotides were infected (MOI 10) with GFP‐producing L. pneumophila JR32 or ΔicmT (pNT28). Intracellular replication was assessed by fluorescence with a microtiter plate reader after 24 h and compared to 1 h. Qiagen AllStars unspecific oligonucleotides (“scrambled”) were used to control for off‐target effects, and an oligonucleotide targeting the small GTPase Arf1 served as positive control. Mean and SEM of three independent experiments (technical sextuplets each) are shown (**P < 0.01, Student's t‐test; Arf1 or Atl3 compared to scrambled).

To test whether Atl3 plays a functional role for intracellular replication of L. pneumophila, the large GTPase was depleted in RAW 264.7 macrophages by RNA interference for 24 h. Depletion of Atl3 was efficient and not more toxic than using control siRNA oligonucleotides (Fig EV5). The macrophages were subsequently infected with GFP‐producing L. pneumophila JR32 or ΔicmT, and intracellular growth was assessed by fluorescence detection. Upon depletion of Atl3, the intracellular growth of L. pneumophila was reduced by approximately 50%, similar to what was observed upon depletion of the small GTPase Arf1 used as a positive control (Fig 7C). Taken together, the large GTPase Atl3 localizes to LCVs in macrophages and promotes intracellular growth of L. pneumophila.

Figure EV5. RNAi‐mediated protein depletion in RAW 264.7 macrophages or A549 cells.

• A, B
  (A) Murine RAW 264.7 macrophages were treated for 24 h with 150 nM siRNA oligonucleotides, or (B) human A549 epithelial cells were treated for 48 h with 10 nM siRNA oligonucleotides, and the efficiency of protein depletion was assessed by Western blot (WB) with the antibodies indicated. For each target protein either four (Atl3, Atl2, Rab10, Rtn4a/b), three (Climp63), or one (Arf1) different oligonucleotides were used (Oligo1–4). Qiagen AllStars unspecific oligonucleotides (“Scrambled”, “Scr”) were used to control for off‐target effects, and GAPDH served as WB loading control. Data are representative of two independent experiments.
• C, D
  The cytotoxicity of the oligonucleotides indicated (Atl3, Atl2, Rab10, Rtn4a/b, Climp63, Arf1) using (C) 150 nM siRNA for RAW 264.7 macrophages after 24 h, or (D) 10 nM siRNA for A549 cells after 48 h was determined by propidium iodide (PI) uptake using flow cytometry. Percentage of PI‐positive cells is shown (means and SD of triplicate experiments). Untreated cells were used as a negative control, and treatment with 70% EtOH for 1 h served as positive control for cell death.

Source data are available online for this figure.

The ER tubule marker Rtn4 localizes to LCVs and promotes growth of Legionella pneumophila

The ER architecture comprises an intricate network of tubules, sheets, and matrices. In RAW 264.7 macrophages, Atl3 colocalizes with calreticulin (Fig 7A) and calnexin (Appendix Fig S4), which also localizes with the ER tubule marker reticulon 4 (Rtn4) and the ER sheet marker Climp63 (cytoskeleton‐linking membrane protein of 63 kDa) (Appendix Fig S4). Rtn4 has been identified by proteomics on purified LCVs from macrophages 24 and was found to associate with the LCV membrane 41.

To analyze in detail the ER structure(s) interacting with LCVs, we compared the localization pattern of the ER tubule markers Atl3 and Rtn4 to the pattern of the ER sheet marker Climp63. RAW 264.7 macrophages were infected with DsRed‐producing L. pneumophila JR32 or ΔicmT, and immuno‐labeled with anti‐Rtn4a/b and anti‐Atl3 antibodies (Fig 8A), or anti‐Rtn4a/b and anti‐Climp63 antibodies (Fig 8C), respectively. At 1 h p.i., approximately 80% of the LCVs were decorated with Rtn4a/b and Atl3 (Fig 8B), compared to only 25% of LCVs positive for Climp63 (Fig 8D). Vacuoles containing the avirulent ΔicmT strain were rarely (< 15%) associated with Rtn4a/b or Atl3.

Figure 8. The ER tubule marker Rtn4 localizes to LCVs and promotes growth of Legionella pneumophila .

• A–D
  RAW 264.7 macrophages were infected (MOI 10, 1 h) with DsRed‐producing L. pneumophila JR32 or ΔicmT (pSW001), and labeled with (A) anti‐Rtn4a/b (reticulon4a/b) and anti‐Atl3 (atlastin3) antibodies, or (C) anti‐Rtn4a/b and anti‐Climp63 (cytoskeleton‐linking membrane protein of 63 kDa) antibodies; scale bars: 10 μm (main image), 1 μm (inset). Quantification of (B) Rtn4a/b‐ and Atl3‐positive LCVs or (D) Rtn4a/b‐ and Climp63‐positive LCVs in RAW 264.7 macrophages at 1 h p.i. Mean and SEM of three independent experiments are shown (n = 50; *P < 0.05, Student's t‐test).
• E, F
  A549 epithelial cells transfected for 48 h with 10 nM siRNA oligonucleotides targeting the mRNAs indicated on the x‐axis were infected (MOI 10) with GFP‐producing L. pneumophila JR32 or ΔicmT (pNT28). Intracellular replication was assessed by fluorescence with a microtiter plate reader after 24 h and compared to 1 h. Qiagen AllStars unspecific oligonucleotides (“scrambled”) were used to control for off‐target effects, and an oligonucleotide targeting the small GTPase Arf1 served as positive control. Mean and SEM of three independent experiments (technical sextuplets each) are shown (*P < 0.05; **P < 0.01, Student's t‐test; all groups compared to scrambled).

To test whether Rtn4 or Climp63 plays a functional role for intracellular replication of L. pneumophila, the components were depleted by single or dual RNA interference. To this end, human A549 lung epithelial cells were used, which compared to macrophages are more efficiently amenable to RNA interference and more robustly endure the transfection procedure. The A549 cells were treated for 48 h with siRNA oligonucleotides targeting Rtn4 or Atl3 alone or simultaneously. This procedure efficiently depleted the target proteins and was not toxic (Fig EV5). The cells were subsequently infected with GFP‐producing L. pneumophila JR32 or ΔicmT, and intracellular growth was assessed by fluorescence detection. Upon depletion of Rtn4, the intracellular growth of L. pneumophila was reduced by approximately 40%, similar to what was observed upon depletion of Atl3 or Arf1 (Fig 8E). The simultaneous depletion of Rtn4 and Atl3 did not further reduce intracellular bacterial growth. Moreover, the depletion of Climp63 (Fig 8E) or Atl2 (Appendix Fig S5) had no effect on the intracellular replication of L. pneumophila.

Similar to Atl3, the small GTPase Rab10 has been implicated in the dynamics and morphology of ER tubules at the leading edge of tubules 42. Accordingly, the depletion of Rab10 or the simultaneous depletion of Atl3 and Rab10 reduced the intracellular growth of L. pneumophila in A549 cells by ca. 40% or 60%, respectively (Fig 8F). In summary, ER tubules rather than ER sheets localize to LCVs in mammalian cells, and Atl3 as well as Rtn4 and Rab10 but not Climp63 promotes intracellular growth of L. pneumophila.

Discussion

The ER tubule‐resident, dynamin‐like large GTPase Sey1/Atl3 mediates homotypic membrane fusion and thus maintains the architecture of the highly dynamic network of ER tubules and sheets. Here, we show that Sey1/Atl3 localizes to LCVs in D. discoideum and macrophages and promotes intracellular replication of L. pneumophila by remodeling the ER connecting with the pathogen vacuole. D. discoideum Sey1 shows GTPase activity and is rendered inactive upon mutation of the conserved P‐loop lysine residue, similar to other atlastin family members 35, 43, 44.

The depletion of Atl3 by RNA interference reduced intracellular replication of L. pneumophila approximately twofold, similar to what has been observed for other host factors positively implicated in growth of the pathogen 25, 45. Contrarily, the depletion of host factors restricting intracellular replication of L. pneumophila increased growth of the pathogen approximately twofold 46. Moreover, while the production of GFP‐Sey1 in D. discoideum promoted intracellular replication of L. pneumophila, GFP‐Sey1_K154A reduced growth, indicating that the mutant protein acts as a dominant‐negative form in the amoebae. Dictyostelium discoideum lacking Sey1 is not available, and the GTPase is likely essential, since the amoebae produce only one atlastin homologue. The relatively small effect of depleting Sey1 (or other host factors) on intracellular replication of L. pneumophila is presumably owing to the robust mechanism of phagocyte resistance and intracellular replication of the pathogen.

In general, the presence or absence of ER markers (Atl3, Rtn4, Climp63) on LCVs positively correlates with an effect on intracellular replication of L. pneumophila: Atl3 and Rtn4 but not Climp63 localize to LCV and promote bacterial replication. However, the ER tubule proteins Atl3 and Rtn4 seem to adopt a specific function during L. pneumophila infection, since the deletion of the calcium‐binding ER proteins calnexin and/or calreticulin, which also localize to LCVs, did not significantly affect intracellular replication of L. pneumophila in D. discoideum 47.

Legionella‐containing vacuole maturation represents a sequential and complex process involving different steps (Fig 9). In a first step, L. pneumophila is taken up by phagocytic host cells through macropinocytosis 47, 48. The PtdIns(3)P‐positive phagosome then undergoes PI conversion to become a PtdIns(4)P‐positive compartment 27. Several L. pneumophila Icm/Dot‐translocated effector proteins, including SidC 18, 47, 49, 50 and SidM 51, 52, anchor to the LCV membrane by selectively binding to PtdIns(4)P (Fig 9). The PtdIns(4)P‐binding effectors promote the recruitment of calnexin‐ and Sey1/Atl3‐positive ER around the LCV, followed by the Sey1/Atl3‐dependent ER remodeling and homotypic fusion. These fusion events wrap the ER tightly around the pathogen vacuole and appear to contribute to the increase of its membrane surface (Fig 9). Intracellular bacterial replication then takes place in an expanding LCV. In summary, our findings are in agreement with the notion that LCVs are formed by a sequential acquisition of PtdIns(3)P, PI conversion to PtdIns(4)P, anchoring of the PtdIns(4)P‐binding effectors SidM and SidC, recruitment of ER membranes, and Sey1/Atl3‐dependent ER remodeling and LCV expansion.

Figure 9. Model of sequential LCV maturation and the role of Sey1/Atl3.

LCV formation and maturation represent a sequential and complex process involving different steps: (I) uptake of L. pneumophila by macropinocytosis, (II) phagosome PI conversion from PtdIns(3)P to PtdIns(4)P, (III) anchoring of Icm/Dot‐translocated effectors (SidC, SidM) to PtdIns(4)P, (IV) recruitment of calnexin‐ and Sey1/Atl3‐positive ER to the LCV, (V) Sey1/Atl3‐dependent ER remodeling and homotypic fusion, and (VI) intracellular bacterial replication and expansion of the LCV.

The accumulation of GFP‐Sey1 to LCVs is evident at 1 h p.i. but not at 30 min p.i. Thus, the acquisition of GFP‐Sey1 seems indeed a rather slow process, corresponding to the notion that Sey1 plays a role in later steps of LCV maturation. Accordingly, the initial accumulation of ER to PtdIns(4)P‐positive LCVs does not seem to require functional Sey1, as the resident ER marker calnexin was recruited to the pathogen vacuole in D. discoideum producing either GFP‐Sey1 or catalytically inactive, likely dominant‐negative GFP‐Sey1_K154A. However, the volume of calnexin‐positive LCV membranes significantly expanded upon overproduction of GFP‐Sey1 in D. discoideum. These results are in agreement with the notion that Sey1 catalyzes the accumulation and fusion of ER‐derived membranes with the ER‐decorated pathogen vacuole. At the same time, Sey1 promoted the expansion of the PtdIns(4)P‐positive limiting LCV membrane, the source or biogenesis of which is currently not known.

The addition of GTP to purified LCVs formed in D. discoideum producing GFP‐Sey1 yielded pathogen vacuole aggregates of twofold larger diameter (due to more individual LCVs) compared to amoebae producing CnxA‐GFP. At this point, we cannot rule out the involvement of GTPase(s) other than Sey1 or (an)other membrane‐bound host factor(s) in this LCV aggregation process. However, due to the extensive washing steps performed during LCV isolation and purification, soluble cytoplasmic host factors are probably diluted and thus seem dispensable for the observed aggregation. Since intact LCVs isolated from D. discoideum producing GFP‐Sey1_K154A are devoid of any aggregation activity, the mutant protein is not only catalytically inactive, but also appears to lack tethering activity.

Interestingly, the overproduction of Sey1 significantly enhanced the intracellular growth of the L. pneumophila parental strain JR32 but not of the ΔsidC‐sdcA mutant strain. This finding is in agreement with the notion that the initial, SidC‐catalyzed attachment of the ER to the LCV precedes and is required for the subsequent function of Sey1. Indeed, Sey1 likely catalyzes later steps of ER remodeling around the LCV, as GTP and the large GTPase promoted the expansion and aggregation of (PtdIns(4)P‐positive) LCVs. It is still poorly understood how the intimate association of the LCV with the ER contributes to intracellular replication of L. pneumophila. The early, PtdIns(4)P‐positive LCV already represents a replication‐permissive compartment, which can contain several bacteria 27. Therefore, later steps during LCV maturation appear to require interactions with the ER.

The intimate interaction and perhaps fusion of the LCV with the ER might provide the source for membrane material required to increase the size of the pathogen vacuole. This notion is in agreement with the findings that Sey1 promotes the expansion of LCVs, and GTP catalyzes the aggregation of intact purified LCVs in vitro. At this point, it is unclear whether and how the PtdIns(4)P‐positive pathogen compartment fuses with the ER. However, the process might involve the PtdIns(4)P‐binding Icm/Dot substrates SidC and SidM, which have been shown to promote the interaction of the LCV with cellular organelles and membranes.

SidM acts as a guanine nucleotide exchange (GEF) for Rab1 and thus recruits the small GTPase to the LCV membrane 53, 54, 55, 56, 57. SidM also covalently modifies the small GTPase by attaching an adenosine monophosphate (AMP) residue 58. This AMPylation causes a prolonged activation of Rab1 and is essential for its recruitment to and retention on LCVs 59. In consequence, the activation of Rab1 by SidM causes the fusion of the LCV with the ER through the non‐canonical pairing of plasma membrane‐derived t‐SNAREs (syntaxin‐2, ‐3, ‐4, SNAP23) with the ER‐resident v‐SNARE Sec22b 22, 60.

SidC is produced by L. pneumophila 61 and by L. longbeachae 49. Legionella spp. that lack the genes encoding SidC and its paralog SdcA form pathogen vacuoles that acquire the ER less efficiently 18, 49. The L. pneumophila ΔsidC‐sdcA mutant grows intracellularly like the parental strain, indicating that ER acquisition is not growth rate limiting. Structural studies of SidC revealed a novel fold of the N‐terminal domain and a catalytic Cys‐His‐Asp triad, which mediates an E3 ubiquitin ligase activity of the effector 62, 63, 64. SidC and its paralog SdcA catalyze the monoubiquitination of Rab1; however, it is unknown how Rab1 ubiquitinylation affects membrane dynamics.

In a complex and only poorly understood manner, LCVs not only communicate with and modulate the secretory vesicle trafficking pathway, but also the endocytic 23, 65, retrograde 46, 66 and autophagy pathways 67, 68, 69. The depletion of Atl3 and Rab10 had a synergistic effect on intracellular replication of L. pneumophila, and therefore, the large and the small GTPase might govern ER dynamics and architecture along the same or different pathways. Rab10 has been identified by proteomics on intact purified LCVs 24. This small GTPase is present on LCVs harboring the virulent L. pneumophila strain JR32, but not on pathogen vacuoles containing ΔicmT. Further studies are required to elucidate the hierarchy and relationship of Atl3 and Rab10 during LCV formation and maturation.

The work described in this paper provides insights into the role of Sey1/Atl3 for the dynamics and architecture of the ER, a crucial cellular organelle, which interacts with numerous membrane compartments. Moreover, our study lays the foundation and makes available tools for a detailed functional analysis of the role of the ER for LCV formation, pathogen vacuole maturation, and intracellular replication of the major human pathogen L. pneumophila.

Materials and Methods

Bacteria, cells, growth conditions, and transformation

Bacterial strains and cell lines used are listed in the Appendix Table S1. Legionella pneumophila strains were grown for 3 days on charcoal yeast extract (CYE) agar plates, buffered with N‐(2‐acetamido)‐2‐aminoethane sulfonic acid (ACES) at 37°C. Liquid cultures in ACES yeast extract (AYE) medium were inoculated at an OD₆₀₀ of 0.1 and grown at 37°C for 21 h to an early stationary phase (2 × 10⁹ bacteria/ml). Chloramphenicol (Cam; 5 μg/ml) was added as required.

Murine macrophage‐like RAW 264.7 cells and human A549 lung epithelial carcinoma cells were cultivated in RPMI 1640 medium (Life Technologies) supplemented with 10% heat‐inactivated fetal bovine serum (FBS; Life Technologies) and 1% glutamine (Life Technologies). The cells were incubated at 37°C with 5% CO₂ in a humidified atmosphere. Cultivation in HL‐5 medium (ForMedium) at 23°C and transformation of axenic D. discoideum Ax3 amoebae were performed as described 27, 70. Transformation was repeated to obtain tandem fluorescent strains. Geneticin (G418, 20 μg/ml) and hygromycin (50 μg/ml) were added as necessary.

Intracellular replication of Legionella pneumophila

Intracellular growth of L. pneumophila JR32, ΔicmT, and ΔsidC‐sdcA in D. discoideum Ax3 was assessed by determining colony‐forming units (CFUs) as well as measuring fluorescence increase during intracellular replication of mCherry‐producing Legionella strains. For assessing CFUs, D. discoideum Ax3 producing GFP, CnxA‐GFP, GFP‐Sey1, or GFP‐Sey1_K154A were seeded at 2 × 10⁴ cells/well in 96‐well cell culture‐treated plates (VWR) and cultured at 23°C in HL‐5 medium. Subsequently, the cells were infected (MOI 10) with mCherry‐producing L. pneumophila JR32, ΔicmT, or ΔsidC‐sdcA, grown for 21 h, diluted in LoFlo medium (ForMedium), centrifuged, and incubated for 1 h at 25°C. The infected cells were washed three times with LoFlo medium and incubated for 24 h (well plate was kept moist with water in extra wells) at 25°C. The amoebae were lysed for 10 min at 25°C with 0.8% saponin (47036, Sigma‐Aldrich), and serial dilutions were plated on CYE agar plates containing Cam (5 μg/ml) and incubated for 3 days at 37°C. CFUs were counted using a colony counter (CounterMat Flash 4000, IUL Instruments, CounterMat software). To determine the intracellular growth of L. pneumophila by fluorescence, mCherry fluorescence was measured using a microtiter plate reader (Synergy H1, BioTek) at 1 h and 24 h post‐infection.

To determine intracellular replication of GFP‐producing L. pneumophila, RAW 264.7 macrophages, A549 cells, or D. discoideum amoebae was infected (MOI 10) with strain JR32, ΔicmT, or ΔsidC‐sdcA harboring plasmid pNT28. The bacteria were grown for 21 h in AYE medium, diluted in RPMI 1640 supplemented with 10% FBS/1% glutamine (mammalian cells) or HL‐5 medium (amoebae), centrifuged (450 g, 10 min, RT), and incubated for 1 h. The infected cells were washed three times with pre‐warmed RPMI 1640 containing 10% FCS or HL‐5 medium and incubated for 24 h or the time indicated (well plate was kept moist with water in extra wells). GFP fluorescence was measured using the plate reader specified above.

Molecular cloning

All plasmids constructed and used are listed in the Appendix Table S1. Translational GFP fusions of Sey1 were constructed by PCR amplification of the putative open reading frame using D. discoideum cDNA (NBRP; Nenkin, Japan) and the primers oBS001_fwd (5′‐AAAAACGCAGATCTAAAATGAGCGAAGAAATAAC‐3′) and oBS002_rev (5′‐TTTTTCGCACTAGTTTCTCTTTTTTGTTTTGG‐3′), respectively. The DNA fragment was cloned into the BglII and SpeI sites of pDM317, yielding pBS001 (GFP‐Sey1).

The GFP‐Sey1 catalytically inactive mutant (Sey1_K154A) was obtained by exchanging the codon AAG (lysine) in position 154 to GCG coding for alanine. Nucleotide substitution was carried out by site‐directed mutagenesis according to the manufacturer's recommendation (QuickChange, Agilent) using pBS001 (GFP‐Sey1) as template and the PAGE purified primers oBS012_fwd (5′‐TGTATTGAATAATAAATTCAATAATGTACTCGCACCACTACTTTGTGGTCCTAAAATTGAAAT‐3′) and oBS013_rev (5′‐ATTTCAATTTTAGGACCACAAAGTAGTGGTGCGAGTACATTATTGAATTTATTATTCAATACA‐3′), yielding pNP108 (Sey1_K154A).

The plasmids pAW12 and pWS032 encoding fusion proteins of red fluorescent mCherry with calnexin or the P4C domain of SidC were constructed by amplifying the D. discoideum cnxA gene or the P4C domain of SidC, respectively, from templates as previously described 27. The PCR fragments obtained were cloned into the BglII and SpeI sites of pDM1044. To generate the plasmid pAW14 for constitutive expression of far red fluorescent mPlum, the gene and its ribosome binding site (RBS) were amplified by PCR from mPlum‐pBAD using the primers oAW30_fwd (5′‐ATATATGAATTCGCTAGATTTAAGAAGGAGATATACATATGGTGAGCAAGGGCGAGGAG‐3′; RBS underlined) and oAW31_rev (5′‐ATATATAAGCTTTTAGGCGCCGGTGGAGTGGC‐3′), and ligated into pNT28 71 cut with EcoRI/HindIII.

To generate pNP99, a pMMB207‐C derivative encoding mCerulean under a constitutive promoter, we amplified the mCerulean ORF from pTSARUd2.1 72 using the primers oNP43_fwd (5′‐ACAGAATTCGCTAGATTTAAGAAGGAGATATACATatggtgagcaagggcgagga‐3′; RBS underlined) and oNP43_rev (5′‐AGCCAAGCTTttacttgtacagctcgtccatgc‐3′) and cloned the DNA fragment into pNT28 using the restriction enzymes EcoRI and HindIII. To generate pNP102, a pMMB207‐C derivative constitutively expressing mCherry, the corresponding ORF was excised from pTSARUd2.4s 72 using the restriction enzymes SacI and HindIII and subcloned into pSW001 73. All PCR products were sequenced.

Recombinant protein production

The GTPase domains of Sey1p (S. cerevisiae), Sey1 (D. discoideum), and Sey1_K154A (D. discoideum) were constructed as follows: the plasmid pGEX‐4T‐3‐sey1p 36 was used as a template to amplify the GTPase domain of Sey1p (aa 1‐498) using the oligonucleotides oBS021_fwd (5′‐TATACGCGGATCCATGGCGGATCGTCCGGCC‐3′) and oBS022_rev (5′‐TATACGCCTCGAGTCAATCGTTCATAATATCATCCC‐3′). pBS001 and pNP108 were used as templates to amplify the GTPase domains of Sey1 and Sey1_K154A (aa 1‐630), using oBS023_fwd (5′‐TATACGCGGATCCATGAGCGAAGAAATAACTACAAATC‐3′) and oBS020_rev (5′‐TATACGCCTCGAGTCAGTAAGTTTTAATCTTTTGCCAC‐3′). The PCR products were digested with BamHI and XhoI and ligated into pGEX‐4T‐3 cut with the same enzymes, resulting in pGEX‐4T‐3‐sey1p_1‐498 (pBS005), pGEX‐4T‐3‐sey1_1‐630, (pBS006) and pGEX‐4T‐3‐sey1_K145A_1‐630 (pBS007).

The GTPase domain–GST fusion proteins were produced in E. coli BL21(DE3) in 4 l LB medium containing 100 μg/ml ampicillin. The cultures were treated with 120 μM IPTG at an OD₆₀₀ of 0.8 and grown overnight at 16°C. The cells were harvested, resuspended in A200 buffer (25 mM HEPES‐KOH, pH 7.4, 200 mM KCl, 2 mM EDTA, 10% glycerol, and 2 mM β‐mercaptoethanol), and lysed in a microfluidizer (Microfluidics, Newton, USA). The membranes were pelleted by centrifugation (45 min, 4°C, 40,000 rpm; rotor 45 Ti, Beckman Coulter). The soluble fraction was incubated with 1 ml glutathione (GSH)‐Sepharose (GE Healthcare) for 3 h and washed with 50 ml A100 buffer (25 mM HEPES‐KOH, pH 7.4, 100 mM KCl, 1 mM EDTA, 10% glycerol, and 2 mM β‐mercaptoethanol). The proteins were eluted from GSH‐Sepharose with A100 and 10 mM of reduced GSH (Roth). The fractions containing protein were pooled, and the purified proteins were concentrated to 1 mg/ml in A100 buffer.

GTPase activity assay

The GTPase activity of the GTPase domains of Sey1p (S. cerevisiae) or Sey1/Sey1_K154A (D. discoideum) was assessed as described 36 by using the phosphate assay kit (E‐6646, EnzChek; Invitrogen) according to the manufacturer's recommendations, except that the reaction buffer was changed to A100. In brief, 5 μM of GST‐Sey1p, GST‐Sey1, or GST‐Sey1_K154A (in buffer A100) was mixed with 2 mM GTP (G8877, Sigma‐Aldrich) to yield a total reaction volume of 200 μl in a 96‐well microtiter plate. The samples were pre‐incubated for 15 min at 37°C, and the reaction was started by the addition of 5 mM MgCl₂ (final concentration). Absorbance at 360 nm was measured at 1‐min intervals for 30 min at 37°C with a plate reader (BioTek) and correlated to a phosphate standard curve. Proteins were heat‐inactivated for 10 min at 95°C as negative controls.

RNA interference, determination of protein depletion efficiency, and cytotoxicity

For the RNA interference experiments, A549 cells were grown in 96‐well plates and treated for 48 h with a final concentration of 10 nM of siRNA oligonucleotides (Appendix Table S2). To this end, the siRNA stock (10 μM) was diluted 1:15 in RNAse‐free water, and 3 μl of diluted siRNA was added per well. Allstars siRNA (Qiagen) was used as a negative control. Subsequently, 24.25 μl RPMI medium without FBS was mixed with 0.75 μl HiPerFect transfection reagent (Qiagen), added to the well, mixed, and incubated for 5–10 min at room temperature (RT). In the meantime, cells were diluted in RPMI medium with 10% FBS, and 175 μl of the diluted cells (2 × 10⁴ cells) was added on top of each siRNA‐HiPerFect transfection complex and incubated for 48 h.

RAW 264.7 macrophages were transfected using the AMAXA Cell Line Nucleofector Kit V (Lonza) according to the manufacturer's recommendations and treated for 24 h with a final concentration of 150 nM of siRNA oligonucleotides (Appendix Table S2). Macrophages or epithelial cells were infected with GFP‐producing L. pneumophila strains, and intracellular replication was determined by fluorescence as described above.

The protein depletion efficiency was assessed by Western blot. A549 cells were grown in 24‐well plates and treated for 48 h with a final concentration of 10 nM of siRNA oligonucleotides (Appendix Table S2). To this end, the siRNA stock (10 μM) was diluted 1:15 in RNAse‐free water, and 9 μl of diluted siRNA was added per well. Allstars siRNA (Qiagen) was used as a negative control. Subsequently, 72.75 μl RPMI medium without FBS was mixed with 2.25 μl HiPerFect transfection reagent (Qiagen), added to the well, mixed, and incubated for 5–10 min at RT. In the meantime, cells were diluted in RPMI medium with 10% FBS, and 525 μl of the diluted cells (6 × 10⁴ cells) was added on top of each siRNA‐HiPerFect transfection complex and incubated for 48 h. RAW 264.7 cells were transfected using the AMAXA Cell Line Nucleofector Kit V (Lonza) according to the manufacturer's recommendations and treated for 24 h with a final concentration of 150 nM of siRNA oligonucleotides (Appendix Table S2).

Protein depletion efficiency was assessed as follows: cells were harvested in ice‐cold PBS and lysed with ice‐cold NP‐40 cell lysis buffer, and cell extracts were subjected to SDS–PAGE. After Western blotting, PVDF membranes were blocked with PBS/3% bovine serum albumin (BSA; Sigma‐Aldrich) for 1 h at RT. Subsequently, specific primary antibodies against Atl3 (ab104262, Abcam), Arf1 (ab58578, Abcam), Atl2 (sc‐109213, SantaCruz), Rab10 (R8906, Sigma‐Aldrich), Climp63/CKAP4 (A302‐256A, Bethyl Laboratories), Rtn4a/b/NOGOa/b (monoclonal mouse antibody 11C7; kind gift of Martin Schwab; 74), or GAPDH (2118, Cell Signalling) were diluted 1:500–1:1,000 in blocking buffer and used to stain the indicated proteins (4°C, overnight). Finally, horse radish peroxidase (HRP)‐conjugated secondary antibodies (GE Healthcare Life Sciences) were diluted 1:2,000 in blocking buffer and incubated (1 h, RT). After extensive washing, the enhanced chemiluminescence (ECL) signal was detected with an ImageQuant LAS4000 (GE Healthcare Life Sciences).

To assess cell viability after siRNA treatment, propidium iodide (PI) uptake was measured. A549 cells or RAW 264.7 macrophages were grown and treated with siRNA oligonucleotides (Appendix Table S2) as described above (protein depletion efficiency). The cells were then harvested in ice‐cold PBS and stained for 10 min with 1 μg/ml PI (Life Technologies) in PBS and subjected to flow cytometry analysis (BD FACS Canto II). Gates were set according to forward/sideward scatter properties, and 10,000 events were collected for each sample. Mean and standard error of mean (SEM) of PI‐positive (PE in FL‐2 channel) cells are shown. Cells treated for 10 min with 70% sterile‐filtered ethanol (EtOH) served as positive control for cell death.

(Live‐cell) fluorescence microscopy

Fluorescence microscopy of (infected) D. discoideum amoebae or RAW 264.7 macrophages was performed as described 18, 47, 66. Briefly, exponentially growing macrophages or amoebae were seeded on sterile coverslips coated with poly‐l‐lysine in 24‐well plates at 2.5 × 10⁵ cells/well in RPMI 1640 (supplemented with 10% FBS and the appropriate antibiotic) or HL‐5 medium, respectively, and let grow overnight. The cells were infected with L. pneumophila strains (MOI 10) as described above, the infection was synchronized by centrifugation (450 g, 10 min, RT), and the infected macrophages or amoebae were incubated at 37°C/5% CO₂ or 25°C, respectively, for the indicated time. The macrophages were then washed with 37°C RPMI 1640, and subsequently, the cells were fixed with 4% paraformaldehyde (PFA; Electron Microscopy Sciences) for 15 min at 37°C. D. discoideum Ax3 amoebae were washed with 4°C HL‐5 and fixed for 30 min at 4°C. Next, cells were permeabilized with PBS/0.1% Triton X‐100 for 10 min and subsequently blocked with PBS/1% BSA (1 h, RT).

For RAW 264.7 macrophages, specific primary antibodies against Atl3 (ab104262, Abcam), Atl2 (sc‐109213, SantaCruz), Climp63/CKAP4 (A302‐257A, Bethyl Laboratories), Rtn4a/b/NOGOa/b (polyclonal rabbit serum; kind gift of Martin Schwab; 75), calnexin (ab31290, Abcam), or calreticulin (ab2907 or 22683, Abcam) were diluted 1:250 in blocking buffer and incubated (1 h, RT). For D. discoideum amoebae, specific primary antibodies against calnexin (270‐390‐2, DSHB, University of Iowa), PtdIns(4)P IgM (Z‐P004, Echelon Biosciences), or SidC 47 were used 1:250 in blocking buffer. Specific secondary antibodies from Life Technologies were used at a concentration of 1:250 (1 h, RT) for indirect immunofluorescence staining. Finally, the coverslips were washed three times with PBS and mounted on glass slides using ProLong Diamond antifade mounting medium with or without DAPI (Thermo Fisher Scientific). The samples were analyzed with a Leica TCS SP5 confocal laser scanning microscope (HCX PL APO CS, objective 63×/1.4–0.60 oil; Leica Microsystems).

Live‐cell imaging was performed as described 27 with D. discoideum harvested from approximately 70% confluent cultures. HL‐5 medium was removed, and cultures were washed with 5 ml LoFlo medium (ForMedium) and resuspended in fresh LoFlo medium. The cells were seeded (300 μl) at a density of 4 × 10⁵/ml in eight‐well μ‐slides (Ibidi). Cells were allowed to adhere for 1 h, after which LoFlo medium was replaced. The microscope stage thermostat was set to between 22°C and 25°C. Upon infection (MOI 5) with early stationary phase cultures of L. pneumophila JR32 harboring pNP99 (mCerulean), the eight‐well μ‐slides were immediately centrifuged (2 min, 500 g). Samples were viewed with a Leica TCS SP5 microscope as outlined above, and captured images were processed and exported with ImageJ software (https://imagej.nih.gov/ij/).

Isolation of LCVs and aggregation assay

To purify intact LCVs from D. discoideum amoebae, a well‐established, two‐step protocol was used 23, 24, 40, 76. Briefly, D. discoideum Ax3 producing calnexin (CnxA)‐GFP, GFP‐Sey1, or GFP‐Sey1_K154A were cultured to approximately 80% confluency in T75 cell culture‐treated flasks (Sigma‐Aldrich) at 23°C in HL‐5 medium, including the appropriate antibiotics. Prior to infection, the supernatant was discarded and at least three flasks of each amoeba strain were infected with mCherry‐producing L. pneumophila JR32 (MOI of 20, diluted in HL‐5 without antibiotics) and incubated at 25°C for 2 h. Cells were washed three times with ice‐cold SorC buffer and finally resuspended in 3 ml of homogenization (HS)‐buffer (20 mM HEPES‐KOH, pH 7.2, 250 mM sucrose, 0.5 mM EGTA). Next, the cells were lysed by nine passages through a ball homogenizer (Isobiotech, http://www.isobiotec.com) using an exclusion size of 8 μm. Homogenates were then blocked for 30 min on ice with 2% FBS and subsequently incubated with an affinity‐purified anti‐SidC antibody (1:3,000) 47, followed by a secondary anti‐rabbit antibody coupled to magnetic beads (130‐048‐602, Miltenyi Biotec). The LCVs were separated in a magnetic field and further purified by a density gradient centrifugation step. Next, the suspensions were equally distributed and centrifuged on top of poly‐l‐lysine‐treated sterile cover slips in a 24‐well cell culture plate. To dilute the high concentration of Histodenz, cells were washed once with ice‐cold HS buffer.

For the aggregation assay, the isolated LCVs were further incubated for 30 min at 30°C with 5 mM GDP (G7127, Sigma‐Aldrich), GTP (G8877, Sigma‐Aldrich), or the non‐hydrolysable analog Gpp(NH)p (Guanosine 5′‐[β,γ‐imido]triphosphate trisodium salt hydrate; G0635, Sigma‐Aldrich) in the presence of 4 mM MgCl₂ (M8266, Sigma‐Aldrich). Preparations were fixed with 4% PFA for 30 min at 4°C and either stained for indirect immunofluorescence with an anti‐rabbit AlexaFluor647 antibody (A31573, Life Technologies), recognizing non‐bound anti‐SidC antibody epitopes, or directly mounted on glass slides using ProLong Diamond antifade mounting medium with or without DAPI (Thermo Fisher Scientific). The samples were analyzed with a Leica TCS SP5 microscope as specified above. The number of individual aggregated LCVs per view field was enumerated using the “Analyze/Analyze Particles” tool of ImageJ software, and the overall aggregate diameter per view field was determined using the “Analyze/Measure” tool of ImageJ. The total number of mCherry‐producing L. pneumophila per view field was enumerated using the “Analyze/Analyze Particles” tool of ImageJ. The diameter of individual PtdIns(4)P‐positive LCVs was determined by measuring the P4C‐mCherry‐positive diameter perpendicular to the longest axis of mCerulean‐producing L. pneumophila.

Imaging flow cytometry

Dictyostelium discoideum Ax3 co‐producing GFP fusions of Sey1 or Sey1_K154A and mCherry fusions of CnxA or P4C were seeded in six‐well plates and infected (MOI 5) with L. pneumophila JR32, ΔicmT, or ΔsidC‐sdcA producing mPlum. At the indicated time points, the infected amoebae were detached from the surface and fixed with 4% PFA. Subsequently, 10,000 events were acquired on an imaging flow cytometer (ImageStreamX MkII; Amnis), and after color compensation, analysis was carried out using the IDEAS 6.2 software (Amnis) largely as described 77. Briefly, upon gating of in‐focus single cells that were positive for both GFP and mCherry, cells having internalized L. pneumophila were gated using the IDEAS internalization wizard. The cells containing exactly one bacterium were then gated based on the feature [Spot Count_Spot(M05, mPlum, Bright, 8.5, 1, 0)_4]. This step was included in order to eliminate cells containing multiple bacteria with varying degrees of colocalization with the fusion protein of interest. The included cells (generally > 1,000 per sample) were finally analyzed for colocalization between mPlum (L. pneumophila) and GFP, as well as between mPlum (L. pneumophila) and mCherry, using the colocalization wizard in IDEAS. This wizard uses the log transformed Pearson's correlation coefficient of the localized bright spots with a radius of three pixels or less in two images, providing an IFC colocalization score for each cell. The mean score for the sample is reported.

Electron microscopy

Samples of L. pneumophila‐infected D. discoideum strains were fixed with 4% PFA and 0.1% glutaraldehyde in 0.1 M cacodylate buffer for 1 h and sequentially treated with 2.5% glutaraldehyde for 30 min, 1% OsO4 for 1 h at 0°C, both in 0.1 M cacodylate buffer and 2% uranyl acetate in H₂O for 1 h at 4°C. Samples were dehydrated in an ethanol series and embedded in Epon/Araldite (Sigma‐Aldrich). Ultrathin (50 nm) sections were contrasted with lead citrate and examined with a CM100 transmission electron microscope (FEI) at an acceleration voltage of 80 kV using an Orius 1000 digital camera (Gatan).

Statistical methods

The two‐sample Student's t‐test was applied assuming unequal variance. Data analysis was performed using GraphPad Prism.

Author contributions

BS, ALS, AW, SW, NP, and AK conceived and conducted experiments and edited the paper. CF and UZ performed the electron microscopy, and RWK provided technical input and help. HH conceived and guided experiments and wrote the paper with input from BS, ALS, AW, SW, NP, and RWK.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Expanded View Figures PDF

Source Data for Expanded View

Review Process File

EMBO Reports (2017) 18: 1817–1836