Life Stage-specific Proteomes of Legionella pneumophila Reveal a Highly Differential Abundance of Virulence-associated Dot/Icm effectors*

Major differences in the transcriptional program underlying the phenotypic switch between exponential and post-exponential growth of Legionella pneumophila were formerly described characterizing important alterations in infection capacity. Additionally, a third state is known where the bacteria transform in a viable but nonculturable state under stress, such as starvation. We here describe phase-related proteomic changes in exponential phase (E), postexponential phase (PE) bacteria, and unculturable microcosms (UNC) containing viable but nonculturable state cells, and identify phase-specific proteins. We present data on different bacterial subproteomes of E and PE, such as soluble whole cell proteins, outer membrane-associated proteins, and extracellular proteins. In total, 1368 different proteins were identified, 922 were quantified and 397 showed differential abundance in E/PE. The quantified subproteomes of soluble whole cell proteins, outer membrane-associated proteins, and extracellular proteins; 841, 55, and 77 proteins, respectively, were visualized in Voronoi treemaps. 95 proteins were quantified exclusively in E, such as cell division proteins MreC, FtsN, FtsA, and ZipA; 33 exclusively in PE, such as motility-related proteins of flagellum biogenesis FlgE, FlgK, and FliA; and 9 exclusively in unculturable microcosms soluble whole cell proteins, such as hypothetical, as well as transport/binding-, and metabolism-related proteins. A high frequency of differentially abundant or phase-exclusive proteins was observed among the 91 quantified effectors of the major virulence-associated protein secretion system Dot/Icm (> 60%). 24 were E-exclusive, such as LepA/B, YlfA, MavG, Lpg2271, and 13 were PE-exclusive, such as RalF, VipD, Lem10. The growth phase-related specific abundance of a subset of Dot/Icm virulence effectors was confirmed by means of Western blotting. We therefore conclude that many effectors are predominantly abundant at either E or PE which suggests their phase specific function. The distinct temporal or spatial presence of such proteins might have important implications for functional assignments in the future or for use as life-stage specific markers for pathogen analysis.

Legionella pneumophila is a Gram-negative bacterium ubiquitously found in water and soil. In their natural environment, the bacteria infect, counteract host defense mechanisms, and intracellularly replicate in protozoa, especially amoebae. When inhaled by humans, L. pneumophila uses similar strategies to propagate in alveolar macrophages leading to Legionnaires disease, a severe pneumonia (1)(2)(3)(4). Upon infection of amoebae and human cells, the intracellular life cycle of L. pneumophila includes at least two distinct stages, a replicative phase, where the bacteria efficiently proliferate, and a transmissive phase, where bacterial replication halts as nutrients become scarce. During transmissive phase, the expression of transmission traits, such as motility, cytotoxicity, accumulation of the storage lipid polyhydroxybutyrate, increased osmotic robustness, production of virulence traits, is triggered and the bacteria are highly infectious (5)(6)(7). The phenotypic switch is regulated in a complex fashion and involves the action of various regulatory proteins, such as CsrA, RpoS, LetA/S, FliA, RelA, and the small ncRNAs RsmY/Z (8 -10). Accordingly, it has been shown that phase transition goes along with significant transcriptomic changes (11)(12)(13)(14) which must impact the life stage-specific proteomes. Remarkably, broth-grown L. pneumophila show a comparable phenotypic switch by expressing a phenotype similar to the replicative phase during exponential phase (E) 1 and to transmissive phase in postexponential phase (PE) (8,12,14).
Currently, several proteomic studies are available for L. pneumophila. One of the first proteomic maps characterizing L. pneumophila grown on agar media was published by Lebeau et al. One hundred ten different proteins of bacterial cell lysates were identified by means of 2D gel electrophoresis (2DE), tryptic digest, and mass spectrometry (MS) (15). Hayashi et al. conducted a similar approach and analyzed the growth phase-dependent L. pneumophila cell extract proteome after growth in broth. They determined four proteins overabundant in E and 64 proteins accumulating in PE. About 60% of the differentially abundant proteins were defined as enzymes and were categorized to: carbohydrate metabolism, amino acid metabolism, and lipid metabolism (16).
In the context of bacterial pathogenicity, exported proteins designed to directly interact with host cells are of particular interest. Several protein secretion systems of L. pneumophila are associated with bacterial virulence, such as the Defect in organelle trafficking/Intracellular multiplication (Dot/Icm) type IVB secretion system (T4BSS), the Legionella type II secretion pathway (Lsp) (T2SS), the type I (Lss), and the Twin-arginine translocation (Tat) systems (15,(17)(18)(19)(20)(21)(22)(23)(24). More than 300 and 25 proteins are translocated into the host cell by the T4BSS or are exported by the T2SS, respectively (18,25). Previous proteomics studies gave an insight into the variety of L. pneumophila proteins present in the culture supernatant and on their dependence on type II secretion, Tat secretion, or export via outer membrane vesicles. In those studies, 2DE/MS analysis was used and identified at least 20 type II-secreted proteins, 20 proteins with differential abundance in wild type and a Tat mutant as well as 181 culture supernatant proteins of which 33 specifically were associated with outer membrane vesicles (20,(25)(26)(27)(28). Furthermore, in a recent study by Hoffmann et al., the proteome of intracellular L. pneumophila 1h post infection was analyzed by means of 1D gel electrophoresis and subsequent liquid chromatography-MS/MS (29). Four hundred thirty-four Legionella proteins were detected in Legionella-containing vacuoles (LCV) purified from RAW 264.7 macrophages, or 944 obtained from D. discoideum including 29 or 60 Dot/Icm effectors, respectively (29).
In addition to E and PE, various studies reported on L. pneumophila in a viable but nonculturable (VBNC) state which might be considered as a third life stage devoted to the survival of unfavorable conditions (30 -34). Bacteria in the VBNC state fail to grow on routine laboratory media on which they would normally grow, but retain insignia of life, such as intact membranes, metabolic activity, transcription, respiration, and some attributes of virulence (35). It has been shown that unculturable L. pneumophila may resuscitate to a fully virulent form by passage through an amoebal host (30,33,34,36) and therefore provide a largely undetected source of infections. Unculturable Legionellae have been observed for example after prolonged incubation in laboratory tap water microcosms at different temperatures, including elevated temperature at 42°C (36), heat shock (50 -70°C) (37), following disinfection (31)(32)(33), and after treatment with heavy metals (38).
Major differences in the transcriptional program underlying the phenotypic switch between E and PE in L. pneumophila, which are important in understanding bacterial virulence, were formerly described (11)(12)(13)(14). This and the study by Hayashi et al. indicated that cell-associated proteomes of E and PE L. pneumophila indeed reveal differences (16). However, the big variety and differential abundance of proteins present at the three L. pneumophila life stages and distinct subproteomes (soluble whole cell, outer membrane-associated, and extracellular) have so far not been comprehensively described. Here, we present a global proteomic analysis of E, PE, and UNC of L. pneumophila and visualize the protein composition and occurring changes in Voronoi treemaps. Our results reveal (1) different life stage-and accordingly virulence-related protein patterns, (2) phase-dependent differential protein abundance including many Dot/Icm effectors, (3) life stage-exclusive proteins, and (4) protein localization of a wide array of proteins.

EXPERIMENTAL PROCEDURES
Experimental Design and Statistical Rationale-To delineate growth phase-related proteomic changes of L. pneumophila, E/PE bacteria and UNC were generated and analyzed by mass spectrometry. E and PE bacteria were assessed for swcp, extr, and omap. Each sample was analyzed in biological triplicates to allow for statistical tests and to improve consistency. Protein identification and quality criteria were very strict throughout the study (see section Proteome Analysis). Only proteins identified in each of three biological replicates were quantified by means of normalized spectral abundance factors (NSAFs) (see section Data Analysis). In order to reduce mainly cytosolic contaminants, quantified extr and omap were assessed for quantitative enrichment of individual proteins compared with swcp, and only enriched proteins were further analyzed (see section Data Analysis of omap and extr). To test for phase related differential abundance of proteins, a t test as implemented in Scaffold (v.4.4.5) was used (p Ͻ 0.05). Ratios of quantity of significantly different proteins were log2 transformed and only those were approved who exceeded 1 or fell below Ϫ1 (49). Within proteins with an apparent phase specific abundance, subfractions termed "on/off proteins" and "phase exclusive proteins" were defined (see section Data Analysis).
Generation of E, PE, and UNC of L. pneumophila-L. pneumophila was routinely grown on buffered charcoal yeast extract (BCYE) agar for 2-3 days at 37°C (39). For growth in liquid laboratory medium, L. pneumophila was inoculated at an OD 660 ϭ 0.2-0.3 and was cultured in buffered yeast extract (BYE) broth at 37°C with continuous shaking at 250 rpm. Bacterial growth was checked by determining the optical density of the culture at wavelength 660 nm (OD 660 ) using a Beckman spectrophotometer DU520 (Beckman Coulter, Brea, CA). E and PE L. pneumophila for proteomics analysis were obtained from liquid cultures. E starter cultures were diluted in BYE broth to an OD 660 ϭ 0.2. E samples were removed after ϳ4.5 h shaking at 37°C. PE samples were obtained at 13.5 h when the bacteria entered stationary growth phase. To confirm that the bacteria reached PE, samples were subjected to Western blotting with antibodies directed against flagellin (kindly provided by Klaus Heuner, Robert Koch-Institut Berlin (40)), a growth phase-related marker of PE L. pneumophila (5). For the generation of L. pneumophila UNC, PE cells were washed two times in sterilized tap water and subsequently inoculated in sterilized tap water to an OD660 of 0.3, corresponding to ϳ2 ϫ 10 8 CFU/ml. The water microcosms were statically incubated at 42°C to induce VBNC bacteria as described elsewhere (41). To study the effect of temperature for efficient generation of VBNC L. pneumophila, samples were incubated at 4°C, 21°C, 42°C, and only the latter condition lead to loss of culturability and maintenance of viability in a reasonable time (ϳ110 days).
Analysis of Unculturable Bacteria-Samples were considered UNC when repeated plating of 1 ml cultures containing 10 8 bacteria on BCYE yielded zero CFU after 5 days of incubation at 37°C. To determine the amount of VBNC and dead bacteria, BacLight Live/Dead staining and the cFDA assay (both purchased from Invitrogen) were used. The BacLight Live/Dead staining kit was used according to the manufacturer's protocol and the stained cells were imaged by epifluorescence microscopy. Additionally, the quantity of esterase-positive bacteria was determined after treatment with 6-carboxyfluoresceindiacetate (cFDA) (42). To this purpose, 1 ml bacterial suspension was incubated with 10 l of a 10 mM cFDA (6-cFDA, Sigma-Aldrich, St. Louis, MO) and 2 l of a 100 mM sodium pyruvate solution for 2 h at 37°C. Subsequently, the bacteria were washed in PBS and esterase-positive bacteria exhibiting green fluorescence were quantified by fluorescence microscopy. Stained bacteria were counted by means of the CellC software (43).
Sample Processing for Proteome Analysis-Soluble whole cell proteins (swcp) were extracted from equal bacterial amounts of E/PE or UNC samples at 4°C. Bacterial cells, pelleted from 150 ml culture (6500 ϫ g, 10 min) were resuspended in 1/10th volume ice-cold Tris-EDTA buffer (TE) and washed twice. Pellets were finally resuspended in 10 ml ice-cold TE and lysed by means of an Emulsiflex C3 (Avestin) using 3 cycles at 15.000 psi. Subsequently, lysates were centrifuged at 20,000 ϫ g (4°C) and cleared lysates were subjected to proteome analysis.
Outer membrane-associated proteins (omap) were extracted as described elsewhere (44). Briefly, 50 ml E or PE cultures were pelleted 4 min at 3200 ϫ g (4°C). Then, bacteria were resuspended in 4 ml ice-cold PBS-PMSF (0.1 mM) and omap were biotinylated by addition of 1.6 mg/ml Sulfo-NHS-SS-Biotin (#21331, Thermo Scientific), followed by an incubation step for 2 h on ice with gentle agitation. Biotinylation was stopped by addition of 50 mM Tris pH 8.0 and the cells were washed four times in PBS-PMSF. Biotinylated samples were solubilized by ultrasonication (4 cycles of 30 s, 50% intensity (Sonoplus, Bandelin, Berlin, Germany)) and the lysate was cleared by centrifugation (18,500 ϫ g, 4°C, 30 min). The cleared lysate was mixed with 500 l neutravidin agarose (#29200, Thermo Fisher Scientific, Waltham, MA) and incubated over night at 4°C in a rotation shaker. Next, protein-agarose complexes were washed six times in cold PBS ϩ Nonidet P-40 (1% v/v). Proteins finally were eluted in reducing SDS-PAGE Laemmli buffer for 60 min on ice.
Extracellular E/PE proteins were precipitated from 250 ml cultures by adding 1/10 volume of 100% (w/v) trichloroacetic acid to cell-free culture supernatants and overnight incubation at 4°C. Next, supernatants were centrifuged 17,700 ϫ g at 4°C for 60 min and the protein pellets were washed six times in ice-cold 96% ethanol and two times in 70% ethanol. The washed pellets finally were dried for 30 min in a rotary evaporator at ambient temperature and subsequently resolubilized in 8 M urea. Each time point and each fraction was generated and analyzed in triplicates.
Proteome Analysis-For proteome analysis, protein concentration of the whole cell and the extracellular protein samples was determined (Roti-Nanoquant, Carl ROTH, Karlsruhe, Germany). Subsequently, the samples were analyzed by liquid chromatography-MS according to the workflow described by Bonn et al. (45). Briefly, 20 g protein (swcp and extr) or the complete sample (omap) were separated by 1D electrophoresis and stained with colloidal Coomassie Brilliant Blue G250 overnight. Subsequently, protein containing gel lanes were cut into 10 equisized pieces and destained by washing at least three times for 15 min with 700 l of gel wash buffer (0.2 M ammonium bicarbonate in 30% (v/v) acetonitrile) at 37°C under vigorous shaking. The destained gel pieces were desiccated in a vacuum centrifuge at 30°C and rehydrated with trypsin solution (2 g of modified trypsin (Promega, Madison, WI) in 1 ml of water) for 15 min. Remaining trypsin solution was removed, and the digest was performed overnight at 37°C. The gel pieces were covered with water, and the peptides were eluted from the gel matrix by immersion of the reaction tube in an ultrasonic bath for 15 min. The supernatant containing the peptides was removed, transferred to a glass vial, and concentrated to a final volume of 10 l in a vacuum centrifuge.
For LC-MS/MS measurements, the tryptic digest was subjected to a reversed-phase column chromatography run on an EASYnLC (Proxeon, Odense, Denmark). Emitter tips for the self-packed columns were prepared by pulling out tips using 100 m i.d. fused silica capillaries with an o.d. of 360 m with a laser puller (P2000, Sutter Instruments, Novato, CA). These emitter tips were packed at Ͼ200 bar using Aeris C18 material (3.6 m). Self-packed columns with a length of 20 cm were used in an open vented one-column setup with a loading volume of 10 l at a flow of 500 nl/min at a maximum of 220 bar and a subsequent flow rate of 300 nl/min. Separation of the peptides was achieved by the application of a binary nonlinear 75-min gradient from 5% to 75% acetonitrile in 0.1% acetic acid. The self-packed columns were mounted in a modified nanoelectrospray ion source with liquid junction of the voltage (2400 V) applied between orifice and emitter tip.
MS and MS/MS data were acquired with the LTQ-Orbitrap (Thermo Fisher). After a survey scan at a resolution of 30,000 in the Orbitrap with activated lockmass correction, the five most abundant precursor ions were selected for fragmentation. Singly charged ions as well as ions without detected charge states were not selected for MS/MS analysis. Collision-induced dissociation (CID) fragmentation was performed for 30 ms with normalized collision energy of 35, and the fragment ions were recorded in the linear ion trap.
Sorcerer-Sequest (v 27, rev 11) was used to create peaklists and to search a L. pneumophila forward-reverse protein sequence database. This Database (5970 entries) included the complete proteome set of L. pneumophila Philadelphia-1 extracted from UniprotKB (2013.04.25) and a set of common laboratory contaminants finally compiled with Scaffold (version Scaffold_4.0.3, Proteome Software Inc., Portland, OR). The search was performed assuming the digestion enzyme trypsin. Data were searched with a fragment ion mass tolerance of 1.00 Da and a search tolerance of 5 ppm for the overview scans. Oxidation of methionine was specified in Sequest as a variable modification and two missed cleavages were allowed. Scaffold (version Scaffold_4.4.5, Proteome Software Inc.) was used to validate MS/MS-based peptide and protein identifications. Proteins were only considered as identified if at least two unique peptides matched solid quality criteria (deltaCn Ͼ 0.1 and XCorr Ͼ 2.2; 3.3; 3.5 for doubly, triply, or higher charged peptides). Peptide quality criteria were very strict to result in a false discovery rate of 0% on protein level in all experiments. False discovery rate was calculated in Scaffold accord-ing to Kä ll et al. (46). Details on identified proteins and peptides are provided in supplemental Table S1.
Data Analysis-In the study only proteins identified with at least two exclusive unique peptides were considered. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium (47) via the PRIDE partner repository with the data set identifier PXD001926 and 10.6019/PXD001926. Only proteins identified in each of three biological replicates of either E, PE, or UNC samples were quantified. Quantification was done by using normalized spectral abundance factors (NSAFs) (48). Differentially abundant proteins were identified by means of t test (p Ͻ 0.05), and were only approved when their corresponding log2 transformed ratio of NSAFs exceeded the thresholds of 1 (E-up) or fell below Ϫ1 (PE-up) (49). Calculations of NSAF and p values were done by means of scaffold (v.4.4.5). Quantified proteins which were not identified in any replicate of the other life stage in the same cell subfraction were defined "on/off proteins." "On-proteins" not identified in any subfraction of any other life stage were defined growth phase exclusive.
Data Analysis of Outer Membrane-associated and Extracellular Proteins-To define outer membrane-associated (omap) and extracellular L. pneumophila proteins (extr), we employed a biotinylation approach (44) and precipitated cell free supernatants from broth cultures, respectively. This led to the identification of 837 potential omap, and 633 potential extr, but both data sets contained a high number of predicted cytosolic proteins (71% in omap, 62% in extr, based on PSORTb 3.0 prediction). To identify proteins quantitatively enriched within the biotinylated cell fraction or the precipitated supernatants, we compared quantitative data obtained in both fractions with the corresponding swcp data by t test (two tailed, type 2). Proteins with significantly higher NSAFs (p Ͻ 0.05) compared with the swcp data were subsequently tested for enrichment by calculating fold-change values (Fc), and those showing FcϾ1.5 compared with swcp were considered enriched by the fractionation approaches. This led to the identification of 55 putative omap, and 77 extr, as well as to a significant depletion of cytosolic contaminants (16% in omap, 10% in extr) in the data sets.
Generation of Voronoi Treemaps-Voronoi treemaps have been proven as a powerful tool for the visualization of large proteomic data, mainly of S. aureus and B. subtilis (48,50). In order to visualize our complex data set, we functionally mapped the quantified L. pneumophila proteome based on its TIGR classification (51). In this way, 724 proteins out of a total of 960 quantified proteins (Table I) were attributed to functions. In order to improve functional annotation of the remainder, functions of the remaining 236 proteins were assessed based on their annotation and prediction (11,52,53). Additionally, known substrates of the type II (Lsp) or the type IVB secretion system (Dot/Icm) were manually incorporated into categories "TII secretion" and "Dot/Icm effector", respectively (20, 54 -56), leaving 92 proteins functionally unassigned. Furthermore, proteins encoded in predicted operons were visualized in clusters (57). Mean NSAF values were assigned to the manually curated L. pneumophila treemap and displayed as protein abundance (as a function area) or as log2 transforms of ratios of protein abundances at two given phases (E/PE).

SDS-PAGE and Western
Blotting-Samples were boiled in reducing SDS-PAGE sample buffer and separated on a 10% SDS-polyacrylamide gel (58). For Western blotting, proteins were transferred to a PVDF membrane (Merck-Millipore, Darmstadt, Germany) and probed with polyclonal anti-flagellin antibody from rabbit (40). After repeated washing, flagellin was detected via anti-rabbit-HRP conjugates, and the blot was developed by means of chemiluminescence (ECL kit, GE Amersham, Little Chalfont, UK).
Introduction of Chromosomal HA-tags and Western Blot Analysis of Dot/Icm Effector Abundance in E and PE L. pneumophila-Hemagglutinin (HA) tags were chromosomally introduced in L. pneumophila JR32 to express N-terminally tagged Dot/Icm effectors for comparison of quantities present in E and PE. The 27bp sequence (TAC-CCATACGATGTTCCAGATTACGCT) was introduced in several individual effector genes by a three step procedure described by Zhao and Meresse for Salmonella (59). First, an ϳ1000bp fragment containing the effector gene start codon in the middle was cloned into pGEMT-Easy (Promega) resulting in plasmids pPA203, pPA205, pPA206, pPA208 to pPA211. Next, the HA sequence was introduced adjacent to the effector start codons by "round-the-horn" site-directed mutagenesis (60) resulting in pPA213, pPA215, pPA216, pPA218 to pPA221 (see supplemental Table S2). Next, the resulting plasmids were amplified by inverse PCR using primers binding upstream to the effector's promoter sequence, and outside of the CDS and promoter sequence of the neighboring gene. The amplified, linearized plasmids were ligated either blunt or via HindIII restriction site to the kanamycin resistance marker gene flanked by FRT sites, which itself was amplified from pKD4 by primers Abp0 and Abp2 or Abp0_HindIII and Abp2_HindIII (61), resulting in pPA223, pPA225, pPA226, pPA228 to pPA231 (see supplemental Table S2). Finally, the mutagenic fragments were amplified from pGEMT-EZ by primers pGEMT_a1f and pGEMT_b1r and were blunt ligated into the allelic exchange vector pLAW344, yielding pPA233, pPA235, pPA236,

TABLE I Overview of identified subproteomes of L. pneumophila exponential phase (E), postexponential phase (PE), and unculturable microcosms (UNC). E/PE-up include only quantified proteins with significantly different abundance between E/PE (t-test, p Ͻ 0.05) and log2 NSAF (E/PE) ratios above ϩ1 or below -1, and include on/off proteins (E-on or PE-on). E/PE-on proteins constitute a subfraction of E/PE-up proteins, and were quantified in the respective cell fraction only in the respective growth phase. UNC-on proteins were only detected in swcp of UNC microcosms. Phase-exclusive proteins constitute a subfraction of E/PE/UNC-on proteins, and include those that were, independent of the cell fraction, solely identified in samples of the respective growth phase
a Percentage of total different proteins found compared to 2943 ORFs corresponding to the L. pneumophila Philadelphia-1 genome NC_002932.5.
pPA238 to pPA241 (see supplemental Table S2) (62). All constructs were sequenced to confirm identity and the lack of errors. Allelic exchange was performed as described elsewhere (62). Allelic exchange was confirmed by PCR with primer HA_f which binds the HA-tag gene sequence, and effector-specific reverse primers localized outside of the homology region used for recombination. Resulting clones were grown comparably to the strains for proteomic analysis (see above) and equal amounts of material from E and PE bacteria were subjected to SDS-PAGE and Western blotting with protein detection via anti HA.11 mAb (Covance).

Overview of L. pneumophila E and PE Proteomes-Be-
cause of their distinctive phenotypes and infection capacity, the aim of our study was to acquire a comprehensive proteomic view on the L. pneumophila life stages and to identify stage-exclusive proteins. For E/PE sample preparation, L. pneumophila was grown in broth until E or PE and enrichment of flagellin FliC (also designated FlaA in some studies) was used to confirm entry into PE (12, 63, 64) ( Fig. 1A and 1B). Our data set of E and PE bacteria comprised proteins from three different bacterial subfractions: (1) soluble whole cell proteins (swcp), (2) outer membrane-associated proteins (omap), and (3) extracellular proteins (extr). A total of 2743 L. pneumophila proteins was identified in the different cell fractions of E and PE corresponding to 1368 different proteins (Table I)  i.e. blue ϭ genetic information processing, dark gray ϭ no function assigned, orange/yellowish ϭ metabolism, red ϭ cellular processes, purple ϭ pathogenesis, pink ϭ cell structure, green ϭ signal transduction. of the annotated theoretical L. pneumophila Philadelphia-1 proteome covering 2943 open reading frames (53). From the E/PE data set, 67% corresponding to 922 different proteins were quantified by using normalized spectral abundance factors (NSAFs) (48) (see experimental procedures). Eight hundred sixty-three proteins were quantified in E and 814 in PE bacteria (Table I). The abundance of 397 proteins showed significant differences between E and PE (t test, p Ͻ 0.05); and included 176 proteins more abundant in E (E-up) and 147 proteins more abundant in PE (PE-up) which were above or below the threshold of ϩ1 and Ϫ1 (log2 NSAF (E/PE)), respectively (see experimental procedures). Both groups contained a high number of on/off proteins, which were, corresponding to a specific cellular subfraction, solely identified either in E or PE. Specifically, a total of 113 proteins was E-on (synonymously PE-off), and a total of 62 was PE-on (synonymously E-off) in swcp, omap and extr. Ninety-five E-on proteins were E-exclusively and 33 PE-on proteins were PE-exclusively identified and quantified (Table I), i.e. were solely detected at either E or PE (and also not in UNC) considering all cellular subfractions. Complete lists of proteins quantified or identified in swcp, and quantified in omap, or extr are given as supplementary data (supplemental Table S3 to S13).
Overview of L. pneumophila E and PE Soluble Whole Cell Proteomes-To comprehensively visualize proteomic data of E and PE, and transition between the phases, we generated Voronoi treemaps (65). Voronoi treemaps have previously been used to combine quantitative proteomic data and functional relatedness of proteins for S. aureus and B. subtilis (48,66). Functional relatedness is represented in hierarchical clusters of irregular tiles and (1) quantitative information is given by the defined size of the respective tile or (2) changing ratios between two conditions are expressed in a color-encoded manner.  Table S3 give an overview about the 841 proteins quantified in E and PE swcp. Proteins identified but not quantified are provided in supplemental Table S4. Very abundant proteins in E and PE included chaperones such as GroEL, GroES, and DnaK; proteins involved in translation, such as elongation factor EF-Tu, Tsf/EF-Ts, FusA/ EF-G; RNA polymerase subunit alpha RpoA; DNA-binding protein HupB; Lpg0689; single strand-binding protein SSB; moreover, more than 20 ribosomal proteins (Rps and Rpl proteins) and others (Fig. 1C, 1D, supplemental Table S3). The three most abundant proteins in both phases were GroEL, RplL, and EF-Tu, in total contributing to around 10% of the entire protein quantity (67)(68)(69). Interestingly, the general pattern of the major (most abundant) swcp proteins did not change substantially with entry into PE, showing that most of the proteome remains stable upon E/PE transition (Fig. 1C, 1D).
Overview of L. pneumophila E and PE Extracellular Proteomes-To characterize the extracellular proteome of L. pneumophila during E and PE, cell-free culture supernatants were precipitated, subjected to MS analysis, and subsequently assessed for enrichment compared with swcp (see Experimental Procedures). We quantified 77 extr of L. pneumophila and 35 of which were E-or PE-up (Fig. 5, supplemental Fig. S5, supplemental Table S8: arranged according to locus tag, supplemental Table S9: arranged according to: log2 NSAF (E/PE)). The eleven most abundant proteins contributed to 53% of the total extracellular protein quantity. Remarkably, eight of those were substrates of the Lsp T2SS (zinc metalloproteinase ProA, aminopeptidase LapA, chitinase ChiA, endoglucanase CelA, tail-fiber-like protein SclB, hypothetical proteins Lpg1832, Lpg0873, Lpg0956), and only two of which (ChiA, Lpg0956) were differentially abundant, specifically PE-up (20,25). In addition, we quantified 10 more T2SS substrates and therefore quantified the almost entire known T2SS secretome (except NttA/Lpg1385 and IcmX/  Table S8, S9). Remarkably, two proteins clearly formed the most abundant extr during E and PE, these were ProA (mean NSAF E ϭ 1.2, PE ϭ 1.7) and LapA (mean NSAF E ϭ 1.0, PE ϭ 1.5), followed by hypothetical protein Lpg0956 (mean NSAF E ϭ 0.2, PE ϭ 1.1, PE-up). Similarly, Lpg0957, encoded adjacent to Lpg0956, belonged to the majorly abundant extr and was also PE-up (mean NSAF E ϭ 0.1, PE ϭ 0.5; log2 NSAF (E/PE) ϭ Ϫ2.1). An overview of the type II-secreted proteins quantified in the study is given in supplemental Table S10.
Growth Phase-dependent Abundance of Dot/Icm Effectors-A high frequency of differentially abundant proteins in E/PE was observed among the Dot/Icm effectors. Of the 86 quantified effectors in the swcp fraction, 55 were E-or PE-up, and of those 37 were on/off proteins. Remarkably, as much as 24 effectors were E-on, whereas 13 were PE-on. The remaining 18 split into 8 E-up and 10 PE-up proteins (Fig. 1, Fig. 3, Fig. 6A, supplemental Table S11). Among the E-up effectors were the eukaryotic-like effector AnkJ which is required for full virulence of L. pneumophila in mice (103), and SidI which has  S3); i.e. blue ϭ genetic information processing, dark gray ϭ no function assigned, orange/yellowish ϭ metabolism, red ϭ cellular processes, purple ϭ pathogenesis, pink ϭ cell structure, green ϭ signal transduction. lethal effects to yeast and mammalian cells because of perturbation of translational machinery (104). E-on effectors included LegS2, a bacterial homologue of the eukaryotic enzyme sphingosine-1-phosphate lyase (SPL) (105), YlfA which is observed on the endoplasmic reticulum-derived replicative vacuole at later stage of a host cell infection (106,107), as well as LepA and LepB promoting nonlytic release of L. pneumophila from protozoa, the latter is a GTPase-activating protein that regulates removal of Rab proteins from membranes (108 -110) (Fig. 6A, supplemental Table S11). The PE-up proteome included SidC, SdcA, SidM/DrrA, and LpnE. SidC has been shown before to accumulate in PE bacteria and, after anchored to phosphatidylinositol-4 phosphate on LCVs, supports efficient intracellular growth and recruitment of endoplasmic reticulum-derived vesicles to the LCV. SdcA is a paralog of SidC with similar functions whose gene is localized directly upstream of sidC (111,112). SidM/DrrA is a Rab1 guanine nucleotide exchange factor (GEF), which regulates the transport of endoplasmic reticulum-derived vesicles and binds to phosphatidylinositol-4 phosphate (109,(113)(114)(115)(116). LpnE binds phosphatidylinositol 3-phosphate, is required for invasion and the establishment of an infection in macrophages, amoebae as well as A/J mice (117, 118) (supplemental Table S11). PE-on Dot/Icm effectors included RalF, a guanine nucleotide exchange factor activating ARF on LCVs (119); VipD, a phospholipase A which blocks endosome fu-sion with LCVs (120, 121); AnkX interfering with fusion of the LCVs with late endosomes (122,123); RavZ inhibiting autophagy during infection (124); and SidH which is negatively regulated by another effector protein, the E3 ubiquitin ligase LubX (112, 125) (supplemental Table S11).
To further address the finding of growth phase-specific abundance of Dot/Icm effectors, we chose seven apparently inversely abundant effectors, specifically RavX, AnkF, MavG, Lpg2271 (E-on); VipD, Lem10, RalF (PE-on) and verified their phase specificity in E or PE. To this purpose, the effectors were chromosomally tagged with the HA-tag sequence (27 bp corresponding to about 1.1 kDa). The HA strains were subsequently broth grown to E or PE, and relative effector abundance was determined by Western blotting with anti HA antibodies. Indeed, HA-VipD (70.4 kDa), HA-Lem10 (67.8 kDa), and HA-RalF (43.3 kDa) solely were detected in PE samples whereas HA-Lpg2271 (25.8 kDa), HA-MavG (52.6 kDa), HA-AnkF (105 kDa), and HA-RavX (40 kDa) almost solely were detected in E samples. Therefore the data obtained by means of Western blotting completely confirmed the proteome data (Fig. 6B). It should be noted that HA-RavX migrated in two independently generated strains as an ϳ50 kDa protein. This is in contrast to its predicted apparent molecular weight of ϳ40 kDa and may suggest post-translational modification.
Generation of UNC for Proteome Analysis-We further aimed to describe the major proteins present in bacteria when  S5); i.e. blue ϭ genetic information processing, dark gray ϭ no function assigned, orange/yellowish ϭ metabolism, red ϭ cellular processes, purple ϭ pathogenesis, pink ϭ cell structure, green ϭ signal transduction.
they are kept prolonged periods in tap water, a situation that is very common for Legionella bacteria. Under those conditions, the bacteria have only limited access to nutrients, switch into a VBNC state, and may persist until they get in contact with a host or nutrient-rich condition (30,34,36). To generate VBNC L. pneumophila, we inoculated the bacteria into sterilized tap water and compared the decrease of CFU at the following temperatures: 4°C, 21°C, and 42°C. Incubation at 42°C efficiently yielded in VBNC bacteria in a time frame of ϳ110 days; a finding that relates well to the results of others (34,36). Microcosms at 4°C or 21°C showed only moderate CFU decrease and therefore were not considered for further analysis (Fig. 7A). Samples with complete loss of culturability at 42°C, in the following termed UNC, were as-sessed for total bacterial counts, the presence of intact viable cells by means of cFDA hydrolysis as a measure of esterase activity and live/dead staining (126 -129). We found that an almost unchanged number of total bacteria contained a fraction of more than 50% intact viable cells as indicated by cFDA hydrolysis (Fig. 7B). Live/dead stain analysis yielded in about 25% clearly membrane-intact cells and comprised a similar fraction of cells with intermediate red/green stains which we here did not rate as membrane intact but still might represent viable cells (Fig. 7B and data not shown).
Proteomic Analysis of Unculturable Cells-A total of 348 L. pneumophila proteins were identified and 223 were quantified in UNC swcp (Fig. 7C, Table I, supplemental Table S12). The 12 most abundant proteins in UNC were in decreasing quan-FIG. 6. Differential abundance of 55 Dot/Icm effectors in E and PE. A, 32 effectors were quantified E-up/on and 23 were PE-up/on in swcp. B, Verification of proteomic results by Western blot analysis of HA-tagged effector proteins. Dot/Icm effector genes vipD, lem10, ralF, lpg2271, mavG, ankF, and ravX were chromosomally tagged with the HA-gene sequence in L. pneumophila. The strains were grown to E and PE and comparable amounts of cell lysates were applied and their phase-specific abundance as indicated by the proteome study was analyzed by means of detection via anti-HA antibodies. The left lane shows the molecular weight standard in kilodaltons. AnkF (105kDa) and RavX were run in a 7.5% SDS-PAGE gel whereas other samples were separated in 10% SDS-PAGE gel. tity Lpg0689, a Dps-like DNA-binding stress protein which interestingly was found abundant in starved E. coli and when complexed with DNA renders it DNase-resistant (130); IcmX, a component of the Dot/Icm transporter (131); Lpg0672, a potential acetoacetate decarboxylase (132); 50S ribosomal protein RplL; co-chaperonin GroES; Icd, the citric acid cycle enzyme isocitrate dehydrogenase (133); a PilE-like type IV pilus assembly protein; entry protein and T4BSS effector LpnE (117,134); Com1, a protein promoting bacterial pathogenicity (Lpg1841) (79); a potential polyhydroxybutyrate dehydrogenase BdhA (135); hypothetical protein Lpg2275; and GroEL, the HSP60 chaperone (136), which, with the exception of the PilE-like protein, all belonged to the highly abundant proteins in PE bacteria. This suggests that most of these proteins are kept during starvation, and may be important for persistence.
Further, four Dot/Icm secretion machinery proteins were quantified in UNC samples, such as IcmS, an adaptor to facilitate effector translocation, IcmE/DotG, a protein of the secretion system central channel spanning inner and outer membranes (137), as well as DotA (85); and 18 Dot/Icm effectors, including LpnE, Ceg28, SidC, RalF, and LidA (all except Ceg28 PE-on or PE-up) (supplemental Table S12) (95, 111, 112, 117, 119, 134, 138 -144). Overall, it is interesting that the UNC proteome is not generally dominated by stress proteins, and rather shows signatures of active life. Indeed, we quantified 55 proteins linked to "genetic information processing" (including protein synthesis) and 85 to "metabolism." The category "cellular processes," which includes stress-related proteins comprises 17 proteins and none of them belongs to the top abundant proteins in these cells. Stress-related proteins in UNC samples included superoxide dismutase, diverse i.e. blue ϭ genetic information processing, dark gray ϭ no function assigned, orange/yellowish ϭ metabolism, red ϭ cellular processes, purple ϭ pathogenesis, pink ϭ cell structure, green ϭ signal transduction. heat shock proteins (small heat shock protein HspC2 Lpg2493, small HspC2 heat shock protein Lpg2192, heat shock protein GrpE, heat shock protein 90/HtpG), catalase/ (hydro)peroxidase KatG, general stress protein GspA, AhpC/ Tsa family peroxynitrite reductase, multidrug resistance protein Lpg0720, as well as Dps-like DNA-binding stress protein Lpg0689. Within the quantified UNC proteome, all but 15 UNC-on proteins overlapped with the identified swcp E/PE proteome.
Definition of Life Stage-exclusive Proteins-Our goal was to define specific proteomic markers of the three distinct L. pneumophila life stages. After removal of all not strictly phaseexclusive proteins, 95 of the 113 different E-on proteins, 33 of the 62 PE-on proteins, and nine of the 15 UNC-on proteins were defined as E-, PE-, or UNC-exclusive proteins (Table I,  Tables II, III, IV).
UNC-exclusive Proteins-Interestingly, only three of the nine UNC-exclusives are functionally annotated proteins: glyoxylase Lpg0619 ("Central intermediary metabolism"), major facilitator transporter Lpg0652 ("Transport and binding"), and PilA/PilE-like Tfp pilus assembly protein ("Cell envelope") ( Table IV). Two hypothetical proteins, Lpg0741 and Lpg2958, were functionally predicted to the category "Energy metabolism" (TIGR00393) and "Protein fate" (TIGR00706), respectively. The further four hypothetical proteins in that fraction do not match any functional prediction. Lpg0619 is annotated as glyoxalase and possesses limited protein homology to the SgaA_N_like (e-value 4.8e-43) and Glycoxalase_2 domains (e-value 4.8e-21), however there is no overall homology to E. coli glyoxalase I and II. Lpg0619 is the L. pneumophila enzyme most closely related to Streptomyces griseus SgaA, a protein involved in Streptomyces growth regulation (150). Lpg0652 bears a major facilitator protein domain (MFS) which comprises a large and diverse group of secondary transporters facilitating transport of various metabolites across membranes (151). Hypothetical protein Lpg0741 carries a CBS domain (Pfam00571), a conserved domain known from cystathionine beta synthase which is frequently found in different bacterial and eukaryotic proteins. The PilA/PilE-like protein Lpg1915, although annotated as PilE, is only distantly related to the previously described PilE protein involved in twitching motility (BlastP: 26% coverage, e-value 6e-07), adherence to cells, and natural competence in L. pneumophila (152)(153)(154). However, its exclusiveness to UNC contradicts the transcriptomic findings of others (155), showing that expression of PilA is affected by sensor kinase LqsS and its homolog LqsT in the stationary phase and that the protein was identified in purified Legionella-containing vacuoles (29). Lpg2958 carries a conserved Clp_protease_NfeD_1 domain (2.31e-94), a domain family which include membrane-bound ClpP class serine proteases acting in protein quality control (156).
To summarize, we analyzed and portrayed the major protein composition of E, PE and UNC of L. pneumophila as well as swcp, omap, and extr of E/PE. We here identified a series of proteins with growth phase-specific abundancies, some of which with even exclusive presence in E, PE, or UNC. Interestingly many Dot/Icm effectors showed differential abundance especially in E and PE which suggests their phasespecific function. The distinct temporal or spatial presence of such proteins might have important implications for functional assignments in the future or for use as life-stage specific markers for pathogen analysis.   (Table I). Such data may be important for selection of significantly abundant proteins for future analysis and may facilitate investigations of protein function and localization.
Three hundred seventy-nine L. pneumophila proteins showed differential abundance in E/PE. 176 of those additionally exceeded the threshold of log2 NSAF (E/PE) ϭ 1 and therefore were considered E-up, whereas 147 fell below the threshold of Ϫ1 and therefore were considered PE-up (Table I). Our data therefore indicate an almost comparable number of proteins differentially abundant in E or PE. This is in contrast to the results of Hayashi et al., who identified in a 2DE-based proteomic study 68 differentially abundant L. pneumophila proteins in E/PE with 64 of which (ϭ 94%) being more abundant in PE than in E (16). Although, several highly abundant proteins associated with flagellum biogenesis were found overabundant in PE samples in our study, especially FliC (mean NSAF PE swcp ϭ 0.31) which was also used as PE marker, the protein strikingly was not among the differentially abundant proteins of Hayashi et al. (Fig. 1, Fig. 5). fliC gene up-regulation is among the highest in amoebae infections with L. pneumophila or during switch from E to PE upon growth of the bacteria in broth (12,14).
The transcriptional study by Brü ggemann et al. revealed that 84% of the up-regulated genes in replicative phase (upon amoebae infection) are also up-regulated during E (upon growth in broth) and 77% of transmissive phase up-regulated genes are also found up-regulated in PE (12). Conformingly, we found that there is a high overlap of the PE-exclusive proteins defined in our study and the transcriptional up-regulation of those genes during transmissive phase in A. castellanii (Ͼ 80%) as well as for the E-exclusive proteins with respect to replicative phase (Ͼ 60%). The more limited representation of the E-exclusive proteins by the replicative phase up-regulated genes might be because of post-transcriptional processes that contribute to reduction of protein abundance not reflected in the transcriptomic data. Brü ggemann et al. further found that some genes are only transcriptionally up-regulated under in vivo infection conditions, including ralF, lidA, and sidC (12). However, in our study, using broth-grown bacteria, the respective proteins were differentially abundant in E/PE swcp. RalF was PE-exclusive, and both LidA and SidC were PE-on proteins (log2 NSAF (E/PE) ϭ Ϫ1,4, and Ϫ2,4, respectively) (supplemental Table S11). Interestingly, RalF, LidA, and SidC were shown before to accumulate to higher protein levels in PE suggesting for some proteins major differences between transcript and protein abundance (112,119).
Our proteome data included a large number of Dot/Icm effectors and some of which showed substantial effects on L.   (Fig. 3, Fig. 6A, 6B, supplemental Table  S11). This implies that Dot/Icm effectors play a substantial role in both E and PE and most of the detected effectors may reveal functions associated with a specific life stage. Interestingly, several effectors known to be important at one specific stage in intracellular lifestyle of the bacteria in our study were most abundant in the other phase. Such as, LepA and LepB, supporting nonlytic release of the bacteria from amoebae, here were more abundant in E (108). Accordingly, YlfA was observed on the endoplasmic reticulum-derived replicative vacuole and on punctate structures at later stages of a host cell infection (106) although its abundance in our study was increased in E. It may therefore be an important strategy of L. pneumophila that some effectors required at later stages of infection already accumulate in E so that they may be right away applicable to the host cell. Correspondingly, some Dot/ Icm effectors, such as SidC, RalF, and LidA were shown before to accumulate to higher protein levels in PE (112,119) and are important at early stages of host infection (157)(158)(159)(160) (111,162), SidM/DrrA (113,115,163) but also some E-up proteins, such as LegS2 (105), LepA (108), and those with equal abundance, such as Lem3 (164), Ceg28/ RidL (165), SdhA (166), SidD (110,167). This shows that a broad variety of Dot/Icm effectors is present in LCV proteomes. Because, so far no quantitative comparison between several time points of the infection has been undertaken; it is not possible to predict whether protein quantities were about to rise or decrease at the specific chosen time point.
So far at least three 2DE and MS-based proteomics studies analyzed extracellular proteins of L. pneumophila (26 -28). Two compared the extracellular proteome of wild type L. pneumophila with secretion mutants, such as the type II or Tat secretion mutants, thus defining secretion system-dependent proteomes (26,27). Indeed in our study, we quantified in the extracellular fraction 20 type II-secreted proteins (two of which have been described as Dot/Icm substrates too, LegY/ GamA, and LegP) and these represented the proteins which were differentially abundant in the T2SS mutant (26,168). Additionally, in extr we quantified 16 of the 20 proteins identified by DeBuck et al. in the supernatant of L. pneumophila wild type and the Tat mutant (27) (supplemental Table S8). Nevertheless, the here presented extracellular proteome contained a subset of proteins which, based on predictions, were unlikely candidates for classic secretion. However, L. pneumophila is known to shed outer membrane vesicles from intact membranes during intra-and extracellular growth (28). The study by Galka et al. analyzed proteins found in outer membrane vesicles expelled by L. pneumophila and culture supernatants devoid of them (28). Seventy percent of our quantified extracellular proteome (54 of 77 proteins) is covered by the secretome data of Galka et al., which is a reasonable overlap and therefore, presence of nonclassically secreted proteins may at least in part be explained by budded outer membrane vesicles. Indeed, eight of the extr in our study which were predicted as periplasmic or membranebound were previously identified in outer membrane vesicles (28). Moreover, six of the predicted 14 cytoplasmic, seven of the 11 predicted membrane, and 14 of the 21 predicted periplasmic proteins in our study were previously identified as L. pneumophila extracellular proteins (28).
Further, export of canonical cytosolic proteins has been described as moonlighting phenomenon. It is however unclear how moonlighting proteins are translocated to the exterior, but their occurrence has been described and is often associated with a shift in function (i.e. in case of GAPDH, enolase) and may result in an increase of bacterial adhesiveness to cells (169). We here used biotinylation of intact L. pneumophila to pull down and characterize omap. This approach has been employed successfully both in Gram-positive and Gram-negative bacteria, and this is the first report in L. pneumophila (170 -174). Biotinylation of proteins was performed by incubation of intact cells with Sulfo-NHS-SS biotin, which is advertised as membrane impermeable and therefore is expected to react specifically with proteins residing in the membrane or being exposed to the exterior (175,176). By means of that method coupled to subsequent enrichment analysis, we detected a variety of known and potentially surface-exposed proteins as well as outer membrane-interacting proteins, such as Mip, EnhA, TolB, OmpH/Lpg0507, and Ttg2D/Lpg0844 (supplemental Table S6) (75,91,177). However, some studies also report on the labeling of intracellular proteins when using Sulfo-NHS-SS biotin (178). Therefore, we further verified the experimentally achieved proteomes by means of enrichment analysis of omap versus swcp data sets. This led to a significant reduction of likely cytosolic contaminants within the omap fraction (70.6% prior enrichment compared with 16.4% after enrichment; according PsortB prediction) and to a significant increase of spatially unpredicted (22.1% prior compared with 40.0% after enrichment) and periplasmic (2.6% to 25.5%) proteins in omap. These results imply that the method not absolutely seems restricted to the detection of proteins located on or in the outer membrane and especially is prone to enrich periplasmic contaminations, likely by biotin leakage through the outer membrane. Nevertheless, the identification/ quantification of a variety of known outer membrane-linked proteins suggests that also several new outer membranelinked proteins were identified in our study. However, their verification needs further dedicated analysis.
Alleron et al. showed by means of an 35 S-labeling approach coupled to subsequent 2D gel electrophoresis-based proteomics, that unculturable L. pneumophila obtained after disinfection with monochloramine actively synthetized proteins and they identified nine proteins which were accumulating in higher quantities in the unculturable bacteria compared with the culturable control (32). Three of those proteins, Mip, DsbA2/Com1, and L1/RplA were quantified in UNC of our study although we used long-term starvation in tap water for generation of UNC.
It was surprising to find IcmX, an essential component with yet unknown specific function in Dot/Icm transport, as the second most abundant UNC protein (mean NSAF 1.26). IcmX has been shown to be essential for bacterial replication in host cells, to be primarily localized in the bacterial periplasm, and a protein fragment was also detected outside of the bacteria (131). However, it is currently not clear which function IcmX may fulfill in the VBNC state.
It is further remarkable that the Com1 protein was found in our study in very prominent amounts in UNC (mean NSAF 0.74) suggesting that this protein plays an important role in this state. It has been described to promote bacterial pathogenicity as well as formation of stable disulfide-bond complexes with substrate proteins including those of the Dot/Icm T4BSS (79).
Finally, our study shows that the phase-exclusive proteomes of E and PE L. pneumophila are in terms of protein numbers clearly dominated by Dot/Icm effectors in conjunction with functionally unassigned and hypothetical proteins. The latter moreover represent several UNC-exclusives. This observation suggests that in large part the mediators attributing for the life stage-related phenotypes of L. pneumophila are unexplored and will require further investigation.