Lipidation by the Host Prenyltransferase Machinery Facilitates Membrane Localization of Legionella pneumophila Effector Proteins

Abstract

The intracellular human pathogen Legionella pneumophila translocates multiple proteins in the host cytosol known as effectors, which subvert host cellular processes to create a membrane-bound organelle that supports bacterial replication. It was observed that several Legionella effectors encode a prototypical eukaryotic prenylation CAAX motif (where C represents a cysteine residue and A denotes an aliphatic amino acid). These bacterial motifs mediated posttranslational modification of effector proteins resulting in the addition of either a farnesyl or geranylgeranyl isoprenyl lipid moiety to the cysteine residue of the CAAX tetrapeptide. Lipidation enhanced membrane affinity for most Legionella CAAX motif proteins and facilitated the localization of these effector proteins to host organelles. Host farnesyltransferase and class I geranylgeranyltransferase were both involved in the lipidation of the Legionella CAAX motif proteins. Perturbation of the host prenylation machinery during infection adversely affected the remodeling of the Legionella-containing vacuole. Thus, these data indicate that Legionella utilize the host prenylation machinery to facilitate targeting of effector proteins to membrane-bound organelles during intracellular infection.

Introduction

When inhaled, the human pathogen Legionella pneumophila can infect and replicate in a specialized vacuolar compartment within alveolar macrophages, causing a severe pneumonia known as Legionnaires' disease (1, 2). Upon phagocytosis, Legionella delays endocytic maturation of the vacuole in which it resides (3, 4) and promotes the recruitment and fusion of early secretory vesicles to the vacuolar membrane (5–7). Ubiquitinated proteins are detected on the mature Legionella-containing vacuole (8, 9), and ER-derived membranes surround and fuse with the Legionella-containing vacuole (10). Bacterial replication commences after biogenesis of this specialized vacuole is completed. Successful establishment of a vacuole that supports bacterial replication is mediated by the collective action of bacterial “effector” proteins (11–13) that are translocated into the host cytosol by Legionella via a specialized type IV secretion system called Dot/Icm (14, 15). The translocated effectors subvert different host processes to conduct vacuole remodeling (16). Mutants lacking a functional Dot/Icm apparatus fail to deliver effectors into the host cytosol and are defective for intracellular replication. Dot/Icm mutants are avirulent in animal models of disease (17–19). Although the functions of most of the over 200 Legionella effectors identified so far remain largely unknown, many of the bacterial effectors have been found to modulate host functions at the cytosolic face of host membrane compartments (20–22). The accumulation of the host small GTPase Rab1 at the Legionella vacuole is controlled by the effector protein DrrA (SidM), which functions as a guanine nucleotide exchange factor (21–24). The protein LepB has GTPase-activating protein activity capable of inactivating Rab1 by promoting GTP hydrolysis (23). How effector proteins target host membranes remains an important question. For the effector proteins SidG, YlfA, and YlfB, there is evidence that hydrophobic domains that insert in membranes mediate protein association with host membranes (25). Recent evidence suggests that specific protein-protein interactions are important for the recruitment of the effector PieA to the Legionella-containing vacuoles (26). Conversely, the effectors DrrA and SidC interact with the lipid phosphatidylinositol 4-phosphate using a domain that is important for host membrane localization (27, 28).

To further define the mechanisms used for membrane localization of Legionella effector proteins, we have been investigating the role of posttranslational modification by host enzymes might play in this process. Incorporation of lipid moiety in a polypeptide increases its hydrophobicity and can result in peripheral association of the modified protein with the membrane lipid bilayer (29). Proteins containing a C-terminal CAAX tetrapeptide motif, where C represents cysteine and A is an aliphatic amino acid, can be lipidated by a class of enzymes known as prenyltransferases. Prenyltransferases use isoprenoid derivatives as their substrates and fall into three categories, depending upon their substrate specificity. Farnesyltransferase (FTase)⁴ and class I geranylgeranyltransferase (GGTase-I), which share a common subunit, append farnesyl (C15) and geranylgeranyl (C20) isoprenoids, respectively, to the cysteine residue of a CAAX motif (30, 31). Conversely, class II GGTase enzyme has been shown primarily to target dicysteine (-CC- or -CXC-) motifs present in the Rab family of small GTPases (32). Prenylated substrates undergo sequential processing by the Rce1 (Ras-converting enzyme 1) endopeptidase and Icmt (isoprenylcysteine-O-carboxyl methyltransferase), which cleave the -AAX tripeptide (33, 34) and methylate the prenylated cysteine residue (35, 36), respectively. Together these modifications facilitate membrane association of prenylated proteins. The best characterized eukaryotic family of CAAX motif-containing proteins is the Ras superfamily of small GTPases, which function as GDP/GTP-controlled switches to regulate diverse cellular processes at the interface of different membrane compartments (37). Although prenylation is required for membrane targeting of Ras GTPases (38–40), it is not sufficient for full membrane association or preferential accumulation at specific membrane organelles. Many CAAX motif proteins encode auxiliary motifs in order to selectively localize at specific cellular compartments (41). HRas and NRas contain upstream cysteine residues that can be sequentially acylated/deacylated to increase/decrease hydrophobicity, altering the subcellular localization (42). Conversely, KRas encodes a series of positively charged residues upstream of the CAAX motif, which creates a charged fold that can interact with acidic groups of lipids embedded in the membrane bilayer, leading to a dynamic range of association with different cellular membranes based on the acidity of the bilayer interface (41, 43, 44). Mutations within these auxiliary domains alter CMP compartmentalization (41). In this study, we demonstrate that subversion of the eukaryotic prenylation machinery is a mechanism that is used for the targeting of Legionella effectors to host membrane-bound compartments.

EXPERIMENTAL PROCEDURES

Bacterial Strains

L. pneumophila serogroup 1, strain Lp01 (45), and the isogenic dotA mutant strain (46) were used in all experiments. Legionella strains were grown on charcoal yeast extract plates (1% yeast extract, 1% N-(2-acetamido)-2-aminoethanesulphonic acid (pH 6.9), 3.3 mm l-cysteine, 0.33 mm Fe(NO3)₃, 1.5% bacto-agar, 0.2% activated charcoal) (47). For all experiments, Legionella were harvested from charcoal yeast extract plates after growth for 2 days at 37 °C.

Cell Culture

HeLa, HK293, and COS-1 cells were cultured in DMEM (Invitrogen), and Raw264.7 cells were cultured in RPMI (Invitrogen) containing 10% fetal bovine serum (Invitrogen) in a humidified tissue culture incubator at 37 °C with 5% CO₂.

Antibodies Used in This Study

polyclonal anti-GFP (Invitrogen), polyclonal anti-calnexin (Stressgen), polyclonal anti-FTα (Santa Cruz Biotechnology), monoclonal anti-GM130 (BD Biosciences), monoclonal anti-ubiquitinated proteins (FK2, Enzo Life Sciences), and monoclonal anti-Lamp1 (1D4B, Developmental Studies Hybridoma Bank, University of Iowa).

Inhibitors

L-744,832 (Calbiochem), FTI-277, GGTI-2133, GGTI-298, and mevastatin (Sigma) were used.

Plasmid Construction

The human HA-HRas and GFP-KRas constructs were kind gifts from Prof. Marilyn Resh and Dr. Mark Philips, respectively. GFP-RhoA was acquired from Addgene (plasmid 12965). Genomic DNA was isolated from L. pneumophila strains Philadelphia, Lens, and Paris. To generate GFP fusion proteins, PCR-amplified bacterial ORFs were digested with BamHI/PstI (lpg2144, lpg2375, and lpg2525), BglII/PstI (lpl2806, lpg1976, lpg0254, and lpg2541), or BglII/KpnI (lpp1863, lpl1858, and lpl2477) and cloned in pEGFP-C2 (Clontech) expression vector digested with BglII/PstI or BglII/KpnI. To generate Cya fusion proteins, bacterial genes were cloned in the bacterial expression vector pEC34 (48) digested with BamHI/PstI. The CS point mutants (Lpg2144 C169S, Lpg2375 C154S, Lpg2525 C441S, Lpl2806 C526S, Lpg1976 C283S, Lpg0254 C186S, Lpl2477 C196S, and Lpg2541 C274S) were generated by incorporating single nucleotide substitutions in the 3′-primer to generate a Cys to Ser mutation. The Lpg2144 K147A/E148A/K149A/K150A (KEKK147AAAA) mutant was generated using the QuikChange mutagenesis methodology (Stratagene). All cloned genes and their mutants were confirmed by sequencing.

Translocation Assay

Translocation of potential substrates into host cells was assayed using the Cya fusion approach. Briefly, HK293 cells expressing FcγRII were plated at 2 × 10⁵ cells/well in a 24-well tissue culture-treated dish and infected the next day with L. pneumophila strains expressing individual Cya-CMP fusion proteins. The bacterial strains were opsonized with anti-Legionella antibody prior to infection (30 min). Infected cultures (MOI = 10) were spun down (5 min, 1000 rpm) to initiate contact and synchronize the infection. Infected cells were incubated for 1 h (37 °C, 5% CO₂) and then washed three times in ice-cold PBS and lysed (30 min, 4 °C) in cold lysis buffer (50 mm HCl, 0.1% Triton X-100). Lysates were boiled (5 min) and neutralized (30 mm NaOH). Cyclic AMP was quantitated using the cAMP Biotrak enzyme immunoassay (GE Healthcare).

Microscopy Analysis of Protein Localization

COS-1 cells were transfected with GFP-tagged CMPs (Fugene-6, Roche Applied Science). Inhibitors were added at 6 h post-transfection and remained in the cell culture for 16 h (L-744,832 (5 μg/ml), GGTI-298 (1 μm), and mevastatin (10 μm)). Cells were fixed (3% paraformaldehyde), washed with PBS, and mounted using ProLong antifade reagent (Invitrogen). Images were acquired with a Zeiss LSM510 microscope using a ×100/1.4 numerical aperture objective.

Subcellular Fractionation

Transfected HK293 1 × 10⁶ cells (Fugene-6, Roche Applied Science) were collected 24 h post-transfection in cold PBS and centrifuged at 1000 rpm for 5 min. To obtain postnuclear supernatant (PNS) fractions, cell pellets were resuspended in 200 μl of hypotonic buffer (150 mm KCl, 20 mm Hepes, 2 mm EDTA, protease inhibitor mixture), passed through a 27-gauge needle (20 times), and centrifuged (3000 rpm, 5 min at 4 °C). The PNS fraction was centrifuged at 100,000 × g for 60 min to obtain the cytosolic fraction (supernatant) and the membrane fraction (pellet). The membrane fraction was resuspended in hypotonic buffer containing 1% Triton X-100 and incubated for 30 min on ice. All fractions were collected in equal buffer volume. Equal volume fractions were resolved by SDS-PAGE, transferred on PVDF membranes, and immunoblotted with different antibodies as indicated.

Differential Extraction of Membrane-associated Proteins

Transfected HK293 (4 × 10⁶) cells were collected 24 h post-transfection in cold PBS and centrifuged at 1000 rpm for 5 min. Cells were lysed in hypotonic buffer by mechanical disruption (27-gauge needle, 20 times) and centrifuged (3000 rpm, 5 min at 4 °C). The volume of the PNS fraction was equilibrated to 400 μl in hypotonic buffer and separated in four 100-μl portions (T in Figs. 3, 4, and 7). Each of the portions was subjected to centrifugation at 100,000 × g for 60 min to separate the cytosol-containing supernatant (C in Figs. 3, 4, and 7) and the membrane-containing pellet fraction. Three of the pellet fractions were resuspended in 100 μl of hypotonic buffer containing either 1% Triton X-100, 1 m NaCl, or 2 m urea and incubated for 30 min on ice. The fourth pellet fraction was resuspended with 0.1 m Na₂CO₃ (pH 11) for 30 min on ice. The solubilized pellet fractions were recentrifuged at 100,000 × g for 60 min to obtain the supernatant containing the soluble membrane fraction (SM in Fig. 4) and the pellet containing the insoluble membrane fraction (PM in Fig. 4). Equal volumes of all fractions were analyzed by SDS-PAGE.

FIGURE 3.

Legionella CMPs associate with membranes. A, representative micrographs of COS-1 cells expressing the indicated Legionella CMP fused to the C terminus of GFP. CS, substitution of cysteine to serine within the CAAX motif. Arrows, perinuclear localization; open triangles, plasma membrane localization; solid triangles, nuclear localization. B and C, PNSs (T) were isolated from HK293 cells expressing the indicated Legionella GFP-CMP fusion proteins or their respective CS mutants and fractionated into a cytosol (C) and a membrane (M) fraction by differential centrifugation. Legionella CMPs were detected in the indicated fractions by immunoblotting (IB) with an anti-GFP antibody. B, relative abundance of each protein within different fractions was quantified using spot densitometry and enumerated as a percentage of combined total. C, as a control for efficient fractionation, the indicated fractions were immunoblotted to detect the transmembrane protein calnexin and the cytosolic protein FTase α (FTα). Scale bar, 10 μm.

FIGURE 4.

Regulation of protein localization and membrane affinity by a Legionella CAAX motif. A, HK293 cells expressing GPF-Lpg2144 or GFP-KRas were treated as indicated with either L-744,832 (FTI) (5 μg/ml) or GGTI-298 (GGTI) (1 μm) or mevastatin (10 μm) for 16 h. PNS (T) was fractionated by differential centrifugation to obtain a cytosol (C) and a membrane (M) fraction. B, PNSs of HK293 cells producing the indicated protein were fractionated by differential centrifugation to obtain cytosol and membrane fractions. Proteins were extracted from the membrane fraction with 1 m NaCl, 2 m urea, 0.1 m Na₂CO₃ (pH 11), or Triton X-100, as indicated, resulting in a membrane-soluble (MS) and a membrane pellet (MP) fraction. Relative abundance of each protein within different fractions was quantified by spot densitometry and enumerated as a percentage of the combined total. C, HK293 cells transfected with the GFP-Lpg2144 C-term truncation containing the C-terminal 44 amino acids of Lpg2144 or the isogenic CS mutant were fractionated and quantified as completed in B. D, as indicated, the fluorescence micrographs of representative COS-1 cells show the localization of GFP-Lpg2144 C-terminal 44 amino acids, the isogenic CS mutant, or C-terminal 33 amino acids after treatment of cells with L-744,832, GGTI-298, and mevastatin. E, schematic of C-terminal region of Lpg2144 with the basic residues highlighted and with the mutated region in the Lpg2144 KEKK147AAAA protein underlined. F, lysates from HK293 cells expressing the GFP-Lpg2144 KEKK147AAAA protein were fractionated and immunoblotted as in A (left), and representative micrographs of COS-1 cells expressing the protein are shown (right). Scale bar, 10 μm.

FIGURE 7.

Prenylation of the CAAX motif facilitates Lpg2144 membrane association in vivo. A, schematic showing how HK293 FcγRII cells infected with Legionella strains expressing a Cya fusion protein (MOI = 10, 4 h) were fractionated. A PNS fraction was prepared in hypotonic buffer using mechanical disruption. Cytosol and membrane fractions were separated by differential centrifugation at 100,000 × g. Endogenous cAMP was depleted from the samples by size exclusion filtration (5 kDa cut-off). The presence of Cya chimera proteins in the different fractions was determined by detecting the cAMP activity generated by the cytosol and membrane fractions in an in vitro cAMP assay. B, expression levels of the indicated Cya fusion proteins in Legionella lysates were determined by M45 immunoblot analysis and show equal levels of production. C, equal volumes of PNS (T), cytosol (C), and membrane (M) fractions were probed for the presence of the trans-membrane FcγRII receptor and FTα subunit as a fractionation control. *, nonspecific band. D, levels of cAMP in the PNS fractions of HK293 FcγRII cells that were infected with Legionella expressing the indicated Cya protein (MOI = 10, 4 h). E, cAMP levels generated in vitro by the cytosol (Cyto) and membrane (Mem) fractions of HK293 FcγRII infected with Legionella strains expressing the indicated Cya protein. Without the addition of exogenous ATP to the reaction, cAMP levels were low (1 × 10² fmol) in all fractions, demonstrating that endogenous cAMP was efficiently removed during fractionation. Shown is the average of triplicate determinations ± S.D. (error bars); the experiment was repeated three times with similar results. F, cAMP levels generated in vitro by the cytosol and membrane fractions of HK293 FcγRII infected with Legionella strains expressing the indicated Cya protein in the presence/absence of L-744,832 (FTI) (5 μg/ml). L-744,832 was added at the time of infection.

Metabolic Labeling with Alk-farnesol-1

HeLa cells were grown in 6-well plates to ∼90% confluence and transfected using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. The analog alk-farnesol-1 was synthesized as detailed by Charron et al.⁵ Cells were treated with alk-farnesol-1 (50 mm) for 4 h using the same volume of DMSO in the negative controls. For prenyltransferase inhibition, HeLa cells were pretreated for 1 h with FTI-277 (10 mm) or GGTI-2133 (10 mm) prior to analog labeling. After labeling, cells were harvested, washed twice with ice-cold PBS, and pelleted at 1000 × g for 5 min. Cells were directly lysed or flash-frozen in liquid nitrogen and stored at −80 °C prior to lysis. Cell pellets were lysed with 150 ml of ice-cold Brij lysis buffer (1% Brij 97, 50 mm triethanolamine, pH 7.4, 150 mm NaCl, 5× EDTA-free Roche protease inhibitor mixture, 10 mm PMSF). Cell lysates were centrifuged at 1000 × g for 5 min at 4 °C to remove cell debris. Tagged proteins were immunoprecipitated from 400 μg of HeLa cell lysate in 250 μl of ice-cold lysis buffer using either 25 μl of packed mouse anti-HA-agarose beads (HA-7, Sigma) or 1 μl of rabbit anti-GFP polyclonal antibody and 50 μl of packed protein A-agarose beads (Roche Applied Science) per sample. Upon incubation at 4 °C for 2 h, the beads were washed with ice-cold modified RIPA lysis buffer (1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS, 50 mm triethanolamine, pH 7.4, 150 mm NaCl).

Cu(I)-catalyzed Huisgen (3 + 2) Cycloaddition/Click Reaction

The immunoprecipitated proteins were suspended in 50 μl of ice-cold PBS, to which was added 3 μl of freshly premixed click chemistry reaction mixture (azido-rhodamine (100 μm, 10 mm stock solution in DMSO), tris(2-carboxyethyl)phosphine hydrochloride (1 mm, 50 mm freshly prepared stock solution in deionized water), tris[(1-benzyl-1H-1,2,3-triazol-4-yl)methyl]amine (100 μm, 10 mm stock solution in DMSO), and CuSO₄·5H₂O (1 mm, 50 mm freshly prepared stock solution in deionized water)) for 1 h at 4 °C on a nutating mixer. The beads were washed (three times, 1 ml each) with ice-cold modified radioimmune precipitation lysis buffer, resuspended in 40 μl of loading buffer (27.5 μl of 4% SDS, 50 mm triethanolamine, pH 7.4, 150 mm NaCl, 10 ml of 4× SDS-loading buffer (40% glycerol, 200 mm Tris-HCl, pH 6.8, 8% SDS, 0.4% bromphenol blue) and 2.5 μl of 0.5 m Bond-Breaker tris(2-carboxyethyl)phosphine hydrochloride Solution (Thermo Scientific)), and heated for 5 min at 95 °C, and 20 μl of the supernatant was loaded on two separate SDS-polyacrylamide gels. One of the gels was used to visualize analog incorporation by in-gel fluorescence imaging, and the other was used to determine total levels of immunoprecipitated proteins by immunoblot analysis using mouse monoclonal anti-HA antibody (HA-7, Sigma) or mouse monoclonal anti-GFP antibody.

In-gel Fluorescence Imaging

SDS-polyacrylamide gels were soaked in destaining solution (40% MeOH, 10% acetic acid) overnight at 4 °C, rehydrated with deionized water, and visualized by scanning the gel on an Amersham Biosciences Typhoon 9400 variable mode imager (excitation 532 nm, 580-nm filter, 30 nm band pass).

Infections of Raw264.7 Cells with L. pneumophila

2 × 10⁵ macrophages were infected with Legionella strains expressing mono-DsRed protein (MOI = 50) for 60 min. Then extracellular bacteria was removed by washing 5 times with warm PBS, and the cells were cultured in RPMI (10% FBS) for an additional 60 min. Medium containing L-744,832 (5 μg/ml) or GGTI-298 (1 μm) or an equivalent volume of DMSO was added at the time of infection. Next, cells were washed with PBS, fixed (3% paraformaldehyde), permeabilized (cold methanol, 30 s), and stained with anti-Lamp1 antibody or anti-ubiquitinated protein antibody. Images were acquired with an inverted microscope (Nikon Eclipse TE2000-S) using a ×100/1.40 numerical aperture oil objective (Nikon Plan Apo) and Hamamatsu ORCA ER camera.

Biochemical Fractionation of Legionella-infected Cells

HK293 cells expressing FcγRII (4 × 10⁶) were infected (MOI = 10) with antibody-opsonized Legionella strains expressing Cya, Cya-Lpg2144, or Cya-Lpg2144 C169S for 2 h. Cells were collected in cold PBS and fractionated as indicated under “Subcellular Fractionation” to generate the PNS, cytosol, and membrane fractions. The final volume from each fraction was 100 μl. The cytosol fraction was passed through PD Spintrap G-25 (5 kDa cut-off) (GE Healthcare) to deplete endogenous cAMP and ATP molecules. 20 μg of the cytosol and the membrane fractions were used to generate cAMP in an in vitro assay.

In Vitro cAMP Assay

The reaction buffer contained 50 mm Tris (pH 8.0), 6 mm MgCl₂·6H₂O, 100 μm CaCl₂, and 500 μg/ml BSA. Isolated cytosol/membrane fractions (20 μl) of infected cells were used in the assay (final reaction volume of 50 μl) in the presence of 2 mm ATP and 100 nm calmodulin (Sigma). The reactions were incubated at 37 °C for 60 min and were terminated by incubating at 96 °C for 10 min. In negative control reactions, ATP was omitted. The levels of cAMP were determined in 10-fold serial dilutions of each in vitro reaction using the cAMP Biotrak enzyme immunoassay (GE Healthcare) against a standard curve.

Quantitation and Statistical Analysis

For quantitation of protein levels in immunoblot analyses, we used the integrated density function of ImageJ (version 1.43i, National Institutes of Health). Scanned TIFF images of individual films were inverted in ImageJ, and the integrated signal density of individual bands was calculated using the circumference of the band as a region of interest. The integrated density represents the product of the area and the mean gray value for the region of interest. Calculations for statistical significance were completed by paired Student's t test (homoscedastic two-tail paired t test) using Excel software (Microsoft).

RESULTS

Identification of Legionella CAAX Motif Proteins

Because Legionella is an intracellular vacuolar pathogen that subverts host functions, we hypothesized that some of the bacterial effectors translocated during infection may utilize host posttranslational modification systems to localize to cellular membranes after translocation. Toward this end, a bioinformatics approach was used to identify Legionella proteins containing a CAAX motif in the C terminus, which might enable effector lipidation by host enzymes. Individual polypeptides from the proteomes of L. pneumophila strain Philadelphia, strain Paris, strain Lens, and strain Corby were analyzed for the presence of a cysteine residue at the fourth position from the C terminus and analyzed using the prenylation prediction program PrePS (49) to identify proteins having an appropriate CAAX motif. Ten Legionella proteins containing high confidence putative CAAX motif were identified (Figs. 1, A and B). Six of these proteins had highly conserved homologs across all Legionella stains, whereas four of the proteins were unique for either the Philadelphia strain or Lens strain (Fig. 1A). Interestingly, the gene lpp2082 encoded a homolog of lpg2144 in which the CAAX motif was deleted, and the homologs lpl1858/lpc_1344 contained a point mutation at the critical cysteine residue (supplemental Fig. 1). The Legionella CMPs lack homology outside of the CAAX motif, suggesting that these proteins encode different functions (Fig. 1C). Lpg2144, Lpg1976, and Lpg1312 are also known as LegAU13/Ceg27/AnkB, LegG1/PieG, and LegC1, respectively.

FIGURE 1.

Identification of a family of CAAX motif-containing polypeptides in the proteome of the intracellular pathogen L. pneumophila. A, distribution of the Legionella CMPs among the four sequenced Legionella strains. Lines connect homologous proteins. Proteins lacking the CAAX motif are highlighted in black. Sequence homology, among the alleles present in the four Legionella strains, was determined using global alignment of protein sequences (BLOSUM62 matrix). B, C terminus amino acid alignment of Legionella CMPs. The shading spectrum (high (black) to low (white)) reflects amino acid similarity. C, putative domains encoded within Legionella CMPs. ANK, ankyrin repeat domain; RCC1, regulator of chromosome condensation domain.

Legionella CMPs Are Dot/Icm Substrates

Because Legionella lacks CAAX motif-specific prenyltransferases, lipidation of the CAAX motif should require translocation of the protein into the cytosol of eukaryotic cells by the Dot/Icm system. Several of the Legionella CMPs have been shown to be substrates of the Dot/Icm system (Lpg2144, Lpg1976, Lpg2525, and Lpg1312) (12, 26, 50, 51). To test whether the other Legionella CMPs are translocated into the host cytosol during infection, we used a well defined enzymatic assay (12, 52). Briefly, the calmodulin-dependent adenylate cyclase domain (Cya) from the Bordetella pertussis CyaA protein was fused to the N terminus of the putative effector protein and expressed in WT Legionella and an isogenic dotA mutant strain (Fig. 2A). The dotA strain lacks a functional Dot/Icm apparatus and is defective in effector protein translocation (20, 52). Delivery of a Cya-effector fusion protein in the host cytosol by the Dot/Icm apparatus activates the Cya enzyme upon binding of calmodulin, resulting in high levels of intracellular cyclic AMP (cAMP) that correlates with the efficiency of translocation of the fusion protein. This process is exemplified by the translocation of the known effector RalF. Cells infected with bacteria producing the Cya-RalF had high levels of cAMP when compared with minimal levels of cAMP detected in cells infected with Legionella expressing Cya alone or dotA mutant strain expressing Cya-RalF or uninfected cells (Fig. 2B). Similar to RalF, cells infected with strains producing Cya-CMP fusion proteins generated 5–200-fold higher levels of intracellular cAMP as compared with control cells infected with dotA strains expressing Cya-CMPs (Fig. 2B), demonstrating that Legionella CMPs are substrates of the Dot/Icm system. These results identify four new Legionella effectors (Lpg2541, Lpg2375, Lpg0254, and Lpl2806) and confirmed translocation of two previously reported effectors (Lpg2144 and Lpg2525) (50). Lpl2477 and Lpp1863 were not tested for translocation because strains expressing the Cya fusion proteins could not be obtained.

FIGURE 2.

Legionella CMPs are substrates of the Dot/Icm system. A, the indicated Cya fusion proteins tagged with the M45 epitope were detected by immunoblot analysis using the M45 monoclonal antibody from Legionella cellular lysates. B, quantitation of total cAMP levels isolated from HK293 FcγRII cells infected with L. pneumophila (MOI = 10) WT (white bars) or dotA (black bars) strains containing plasmids expressing the indicated Cya-effector fusion proteins. UN, uninfected cells; Cya, cells infected with bacteria expressing only the Cya protein. Each bar represents the mean cAMP value from triplicate infections ± S.D. (error bars).

Legionella CAAX Motif Proteins Localize to Membranes

To determine the subcellular location of the Legionella CMPs, GFP fusions were ectopically expressed in COS-1 cells, and their localization was examined by fluorescence microscopy (Fig. 3A). Unlike GFP, which distributed uniformly throughout the cell, the CAAX motif proteins exhibited two distinct subcellular localizations, at the plasma membrane (Lpg2144 and Lpg2541) and at the perinuclear region (the remaining CMPs) (Fig. 3A). Lpg1312 failed to express in eukaryotic cells. The CMPs that accumulated at the perinuclear region co-localized with Golgi proteins, suggesting that they reside at the Golgi compartment (supplemental Fig. 2A). The distinct membrane localization of the Legionella CMPs suggests that these proteins have the capacity to engage membranes. Biochemical fractionation of cells expressing the Legionella CMPs confirmed that these proteins were present in the membrane fraction, which contained the transmembrane protein calnexin (Fig. 3, B and C). These observations demonstrate that the Legionella CMPs encode specific subcellular membrane targeting information within their polypeptide sequence.

CAAX Motif Is Required for Membrane Localization

To test whether the CAAX motif is important for the localization of Legionella CMPs to host membranes, the lipid acceptor cysteine residue within the CAAX motif was mutated to a serine. The lipidation CS mutants either maintained (Lpg2525, Lpl2806, and Lpp1863) or completely lost (Lpg0254, Lpg1976, Lpg2144, Lpg2541, Lpl2477, and Lpg2375) the original subcellular localization of the WT proteins (Fig. 3A). Some lipidation mutants also accumulated in the nucleus. The cytosolic phenotype observed by fluorescence microscopy for the CS mutants was validated using cell fractionation studies, which showed that the mutant proteins are no longer enriched in the cell membrane fractions (Fig. 3B). The majority of the CS mutant proteins were found in the cytosolic fraction, where the control cytosolic protein FTase α was located (Fig. 3, B and C). One exception was the Lpl2806CS protein, which showed significant redistribution into the cytosolic fraction as compared with Lpg2806, but perinuclear localization remained evident by fluorescence microscopy. Thus, the CAAX motif represents a determinant in the CMP proteins that is important for membrane localization.

To further investigate the mechanism by which the CAAX motif of the Legionella CMPs facilitates association with host membranes, we focused specifically on Lpg2144 as a model protein. To address whether lipidation of the CAAX motif facilitates membrane association by a process requiring host lipidation, mevastatin was used as a pharmacological inhibitor of isoprenoid biosynthesis (53). Treatment with mevastatin resulted in redistribution of Lpg2144 from the membrane fraction into the cytosolic fraction (Fig. 4A). Similar results were obtained for the control protein KRas, which is a small GTPase that is anchored to the plasma membrane by a process that requires farnesylation of a C-terminal CAAX motif (Fig. 4A). The functions of the enzymes that lipidate CAAX motifs were also perturbed using the peptidomimetic inhibitors L-744,832 (FTase inhibitor) and GGTI-298 (GGTase-I inhibitor). Similar to mevastatin, L-744,832 increased the cytosolic pool and decreased the membrane-associated pool of Lpg2144. Conversely, GGTI-298 did not alter membrane association of Lpg2144, suggesting that lipidation of the CAAX motif is required to generate membrane affinity and specifically farnesylation mediates membrane association (Fig. 4A).

To determine biochemically whether Lpg2144 associates with membranes using a mechanism that is similar to KRas, proteins were extracted from membrane fractions sequentially using high salt (1 m NaCl), mild denaturing (2 m urea), high pH (0.1 m Na₂CO₃ (pH 11.5)), or detergent (Triton X-100) conditions. Similar to KRas, Lpg2144 was extracted from the membrane fraction using high pH or detergent, which are conditions that perturb the lipid bilayer and extract peripheral as well as integral membrane proteins (54). Unlike the peripheral membrane protein GM130, which associates with membranes through a protein-protein interaction with the lipidated protein GRASP65 (55), Lpg2144 did not extract from the membrane in the presence of 2 m urea, suggesting that Lpg2144 directly interacts with the lipid bilayer. Mutation of Cys to Ser in the CAAX motif of Lpg2144 decreased the protein interaction with the membrane fraction, and the mutant protein was extracted under all conditions, suggesting that lipidation of the CAAX motif of Lpg2144 promotes stable peripheral association of the protein with membranes.

To define the minimum region in Lpg2144 sufficient for membrane targeting, C-terminal truncation mutants were generated by fusing the last 33 and 44 amino acids of Lpg2144 to GFP. The resulting proteins were assayed for their capacity to associate with membranes. Attachment of the last 44 amino acids from the C-terminal region of Lpg2144 to GFP resulted in the redistribution of fusion protein from the cytosol to membranes (Fig. 4, C and D). Conversely, GFP remained cytosolic and distributed throughout the cell when a 33-amino acid C-terminal region of Lpg2144 was used (Fig. 4D and supplemental Fig. 2B). Mutation of the cysteine residue in the CAAX motif to serine eliminated membrane localization of the GFP fusion protein containing the C-terminal domain of Lpg2144 (Fig. 4, C and D). Thus, lipidation of the CAAX motif within the domain is required but not sufficient for membrane localization. In addition to the CAAX motif, the minimal membrane-targeting domain of Lpg2144 encoded 10 positively charged amino acids (Fig. 4E). Stretches of positively charged residues can facilitate ionic interaction with the negatively charged groups present on lipids moieties incorporated in the outer leaflet of the lipid bilayer and thus enhance protein association with membranes, as has been shown for KRas (41). To determine if the C-terminal basic residues are important for Lpg2144 localization to membranes, three of the central lysine residues within the membrane-targeting domain were mutated to alanine. Although the Lpg2144 KEKK147AAAA mutant still partitioned into the membrane fraction as efficiently as the WT protein, fluorescence microscopy indicated that the protein now localized primarily to the Golgi apparatus (Fig. 4F) unlike the WT protein, which resides at the plasma membrane (Fig. 3A). These results indicate that accumulation of the polycationic protein Lpg2144 at the plasma membrane, the most negatively charged membrane interface in the cell (43), requires a lipidated CAAX motif to facilitate membrane association and polybasic residues to provide ionic charge.

Legionella CMPs Are Substrates of Eukaryotic Prenyltransferases

Because the CAAX motif is required for association of the Legionella CMPs proteins with host membranes, we sought to determine which host lipidation enzymes were involved in the presumed posttranslational modifications of these bacterial effectors. Inhibition of isoprenoid biosynthesis with mevastatin resulted in cytosolic or nuclear redistribution of most (Lpg2144, Lpg0254, Lpg1976, Lpg2375, Lpg2541, and Lpl2477) but not all Legionella CMPs (Fig. 5). These results phenocopy the relocalization effects observed for the CS mutants (Fig. 3A), demonstrating that prenylation of the CAAX motif dictates the subcellular localization of the Legionella CMPs. To distinguish which prenyltransferases mediate lipidation of the individual proteins, the activity of FTase or GGTase-I was perturbed using different pharmacological inhibitors, and localization of the bacterial CMPs was determined. Substrates of FTase (Lpg2144 and Lpl2477) redistributed only in the presence of farnesyltransferase inhibitor, whereas substrates of GGTase-I (Lpg1976 and Lpg2375) mislocalized only in the presence of geranylgeranyltransferase inhibitor. Contrary to mevastatin treatment, neither farnesyltransferase inhibitor nor geranylgeranyltransferase inhibitor perturbed the localization of Lpg0254 and Lpg2541, suggesting that FTase and GGTase-I are both capable of modifying these proteins.

FIGURE 5.

Legionella CMPs are substrates of eukaryotic prenyltransferases. Fluorescence micrographs of COS-1 cells producing the indicated GFP-CMP proteins that were treated with L-744,832 (FTI) (5 μg/ml) or GGTI-298 (GGTI) (1 μm) or mevastatin (10 μm) for 16 h as indicated. Arrows, perinuclear localization; open triangles, plasma membrane localization; solid triangles, nuclear protein localization. Scale bar, 10 μm.

To directly determine whether Legionella CMPs are lipidated, Bioorthogonal ligation reactions were conducted using Cu(I)-catalyzed azide-alkyne cycloadditions to monitor CMP incorporation of alkyne-functionalized isoprenol reporter (Fig. 6). This approach has been used to provide rapid and sensitive detection of acylated and prenylated proteins (56–58). The alkynyl-farnesol reporter alk-farnesol-1 (C15) (Fig. 6A) is a substrate for FTase, GGTase-I, and GGTase-II in cells and metabolically installed onto prenylated proteins.⁵ Selective inhibition of prenyltransferases in combination with alk-farnesol-1 metabolic labeling can reveal whether a polypeptide is a substrate for FTase (HRas) or FTase-I/GGTase-I (RhoA) (Fig. 6B). Unlike their CS mutants, Legionella CMPs incorporated alk-farnesol-1 (Fig. 6C), demonstrating that these proteins are prenylated at the Cys residue of the CAAX motif. Lpg2144, Lpl2806, and Lpl2477 were found to be FTase substrates, whereas Lpg0254, Lpg1976, Lpg2375, Lpg2525, Lpl1863, and Lpg2541 were substrates recognized by both FTase and GGTase-I. Lpp1863 failed to incorporate alk-farnesol-1, suggesting that Lpp1863 contains a “non-functional” CAAX motif. These data provide biochemical evidence that Legionella CMPs are lipidated and specific substrates of the eukaryotic prenylation machinery.

FIGURE 6.

Prenylation of Legionella CMPs. A, chemical structure of the alk-farnesol-1 analog with the alkynyl group highlighted. B, incorporation of the alk-farnesol-1 analog (Probe) is shown by direct in-gel fluorescence imaging for cells producing either HRas or RhoA and cultured in the presence or absence (Mock) of FTI-227 (FTI) (10 μm) or GGTI-2133 (GGTI) (10 μm) as indicated. Immunoblot analysis (IB) of total protein serves as a protein loading control. C, alk-farnesol-1 incorporation by Legionella CMPs fusion proteins and isogenic CS mutants in the presence/absence of FTI-227 (10 μm) or GGTI-2133 (10 μm) as indicated. Total protein levels were determined by immunoblot (IB) analysis with anti-GFP antibody. The numbers below each lane indicate relative levels of analog incorporation as determined by spot densitometry.

Host Prenylation Is Important for Membrane Localization of CMPs Translocated during Infection and for Successful Trafficking of Legionella in Host Cells

Because effector proteins are translocated in very low amounts during Legionella infection, it is difficult to monitor their presence and distribution in cells by standard fractionation methods. Thus, to measure the ability of the Legionella CMPs to localize to membranes after translocation into the host cell during infection, we took advantage of a CMP fusion protein having adenylate cyclase appended to the N terminus. This approach allowed enzymatic activity to be monitored after cellular fractionation (Fig. 7A). Cells were infected with Legionella strains expressing either Cya-Lpg2144, Cya-Lpg2144 C169S, or Cya alone (Fig. 7B), and cytosolic and membrane fraction were obtained. The transmembrane FcγRII receptor and cytosolic FTα proteins were used to monitor efficient fractionation (Fig. 7C). Localization of the translocated CMP was determined by measuring adenylate cyclase activity in each fraction after host cells were infected. It was determined that both WT Lpg2144 and the isogenic CS mutant protein were translocated at equivalent levels during infection (Fig. 7D). Upon cellular fractionation, adenylate cyclase activity localized primarily to the membrane fraction in cells infected with Legionella, producing WT Lpg2144 fused to Cya, whereas adenylate cyclase activity localized primarily to the cytosolic fraction in cells infected with Legionella, producing Cya fused to Lpg2144 C169S (Fig. 7E). Importantly, in cells infected with Legionella expressing Cya alone, a situation in which the enzyme is not translocated into host cells, neither the membrane nor the cytosolic fraction contained high levels of adenylate cyclase activity, indicating that the activity of the translocated CMP fusion proteins greatly exceeds the activity of endogenous host enzymes or the activity of Cya that remains inside of bacterial cells. Moreover, inhibition of the host FTase during infection also redistributed Lpg2144 from the membrane to the cytosolic fraction (Fig. 7F). Thus, the CAAX motif in Lpg2144 is an essential determinant that enables this effector protein to associate with membranes after translocation into host cells during Legionella infection.

Successful remodeling of the Legionella vacuole involves inhibition of fusion with late endocytic/lysosomal compartments of the host cell (18, 59). Failure to block endocytic maturation of the vacuole containing Legionella, as illustrated by the dotA mutant, results in acquisition of the host protein Lamp1 (Figs. 8, A and B). Most vacuoles containing WT Legionella are Lamp1-negative because they avoid endocytic maturation. The presence of ubiquitinated proteins on the vacuole and the absence of Lamp1 are indicators of successful trafficking of Legionella in the host cell. (Fig. 8, B and C) (8).

FIGURE 8.

Host prenyltransferases facilitate efficient Legionella vacuole remodeling. A, fluorescence micrographs of Raw264.7 cells infected with WT or dotA mutant Legionella (MOI = 50, 2 h) and stained with α-Lamp1 and α-Legionella antibodies. Bacteria that co-localize with Lamp1 reside within a late endosome/lysosome (arrows). Triangles, bacteria residing within a cellular compartment devoid of Lamp1. B and C, quantitation of Legionella residing in Lamp1-positive (B) or ubiquitin-rich (C) compartments at 2 h postinfection in Raw264.7 cells. DMSO, L-744,832 (FTI) (5 μg/ml), and GGTI-298 (GGTI) (1 μm) were added at the time of infection. The graph represents the average of triplicate determinations ± S.D. (error bars); at least 200 vacuoles were scored. Significant differences were observed in inhibitor versus mock treatment (homoscedastic two-tailed paired t test). N.D, not detected; scale bar, 5 μm.

To test whether interfering with lipidation of the CMPs during infection would influence biogenesis of the vacuole in which Legionella resides, host cells were treated with peptidomimetic inhibitors targeting FTase and GGTase-I, and vacuole remodeling was assessed by determining the frequency by which these compartments avoided Lamp1 acquisition and became decorated with host ubiquitin (Fig. 8, B and C). Perturbation of host prenyltransferases resulted in a significant increase in the percentage of vacuoles that acquired Lamp1 and a decrease in the percentage of vacuoles that stained positive for ubiquitin (75% versus 50%) (Figs. 7, B and C). The defect in vacuole maturation was observed when FTase or GGTase-I was inhibited, and blocking both enzymes simultaneously had no additive effect. Importantly, the prenyltransferase inhibitors did not affect the uptake or trafficking of vacuoles containing dotA mutant bacteria. Collectively, these data indicate that the host protein prenylation machinery enhances the ability of Legionella to remodel its vacuole.

DISCUSSION

Here we show that L. pneumophila encodes a family of effectors containing the CAAX motif that use the host lipidation machinery to mediate membrane localization. Experimental evidence indicates that at least eight different effectors require the CAAX motif for membrane localization because this region mediates lipidation by host prenyltransferases. Using specific prenylation inhibitors, it was determined that some of the Legionella CMPs showed specificity for either FTase or GGTase-I, whereas several of the CMP effectors could be modified by either enzyme (Table 1). The PrePS program predicted that the promiscuous Legionella CAAX motifs could structurally accommodate both enzymes (Table 1). Dual enzyme modification for eukaryotic CMPs substrates is rare; however, under certain circumstances, KRas and HRas can undergo lipidation by both FTase and GGTase-I (60, 61).

TABLE 1.

Legionella CAAX motif substrate specificity

Shown is a comparison between PrePS in silico prediction and experimental results. FTI, farnesyltransferase inhibitor; GGTI, geranylgeranyltransferase inhibitor; NA, not applicable; NC, analysis not completed.

	PrePS substrate prediction		Sensitivity to prenylation inhibitors
	FTase	GGTase-I	Subcellular localization			Analogue incorporation
			FTI	GGTI	Mevastatin	FTI	GGTI
Lpg0254	+	+++	No	No	Yes	Partial	Not affected
Lpg1976	−	+	No	Yes	Yes	Complete	Complete
Lpg2144	+	− −	Yes	No	Yes	Complete	Not affected
Lpg2375	++	+++	Yes	No	Yes	Partial	Partial
Lpg2525	++	+++	No	No	No	Partial	Partial
Lpg2541	++	+++	No	No	Yes	Partial	Not affected
Lpl2477	+++	+++	Yes	No	Yes	Complete	Not affected
Lpl2806	++	− −	No	No	No	Complete	Not affected
Lpp1863	+	+	No	No	No	NA	NA
Lpg1312	++	++	NC	NC	NC	NC	NC

Given that the CAAX motif is a general membrane-targeting determinant, acquisition of this region by a cytosolic protein would enhance membrane affinity. Likewise, a polypeptide that already contains a membrane-targeting domain could significantly strengthen membrane affinity by gaining a CAAX motif. In either case, the CAAX motif could alter protein localization and therefore affect protein function. For example, mutation of the KRas CAAX motif abolished its capacity to transform cells (41). Legionella CMPs exhibited differential dependence on the CAAX motif for localization and membrane affinity. For most Legionella CMPs, localization was linked to the presence of a CAAX motif. The proteins Lpl2806 and Lpg2525 were exceptions in that these proteins retained perinuclear localization in the absence of a functional CAAX motif, which suggests that Lpl2806 and Lpg2525 encode additional membrane-targeting domains, and for these proteins the CAAX motif could play a role in regulating function and not localization. The extent and kinetics of postprenylation processing might also contribute to the differential localization of some Legionella CMPs.

During infection, translocated Lpg2144 preferentially associated with host membranes by a process that required both an intact CAAX motif and functional host FTase. In vivo Lpg2144CS was still detected in the membrane fraction, albeit at lower levels compared with the WT protein, consistent with the idea that prenylation is required for efficient membrane anchoring and subcellular localization, but auxiliary membrane-targeting domain(s) could provide some affinity for membranes presumably by interacting with membrane-resident proteins or lipids.

In eukaryotic proteins, the CAAX motif frequently appears in tandem with either an upstream stretch of basic amino acids (KRas) or cysteine residues (HRas), which can be acylated (41). Although Legionella CMPs lack CAAX motif-proximal cysteine residue, several proteins contain a polycationic amino acid stretch (Lpg2541, Lpg2144, and Lpg1976). For these polypeptides, the number of basic residues in the C terminus correlated with protein localization. Lpg2144 and Lpg2541, which have 10 and 8 basic residues near the C terminus, respectively, localized at the plasma membrane. By contrast, Lpg1976 containing only 6 basic residues localized at the Golgi, consistent with protein charge being an important determinant for organelle distribution of the membrane-localized protein. Indeed, reducing the number of basic amino acids from 10 to 7 in Lpg2144 did not perturb membrane localization but did redistribute the protein from the plasma membrane to the Golgi apparatus, demonstrating that multiple determinants influence membrane distribution patterns. Lipidation-dependent membrane targeting of bacterial effectors is rare and was first discovered in the plant pathogen Pseudomonas syringae cysteine protease effector AvrPphB, which contains an internal myristoylation and acylation motif (62). Upon translocation in the plant cell, AvrPphB undergoes a self-cleavage reaction, revealing the dual lipidation motif that is modified by the host lipid-transferases facilitating plasma membrane localization (62, 63). Additionally, the Salmonella effector SifA has been shown to have a CAAX motif. SifA is both geranylgeranylated and palmitoylated to facilitate membrane association, and a point mutant lacking both lipidation modifications was predominantly cytosolic (64). Therefore, subversion of host lipidation by bacterial pathogens could represent a conserved virulence strategy. This idea is supported by the observation that host FTase and GGTase-I activities are required for correct trafficking of vacuoles containing Legionella, presumably because these enzymes are facilitating the correct targeting of bacterial CMPs to the appropriate host membranes where they exert their effector functions. Because prenylation is a permanent posttranslational modification, depletion of prenylated polypeptides is inherently linked to protein turnover and stability. Therefore, short lived inhibition of prenylation initiated at the time of infection would have far more drastic effect on lipidation of the translocated bacterial effectors rather then the more abundant pools of host prenylated proteins. However, the possibility that short lived prenylated host proteins may contribute to remodeling of the Legionella vacuole cannot be excluded.

In this study, we have identified and characterized a family of Legionella CAAX motif-containing proteins that utilize the host prenylation machinery to associate with defined host subcellular organelles. These data further expand the number of identified substrates of the Dot/Icm apparatus and reveal a new paradigm in targeting Legionella effector proteins to membrane-bound organelles.