Amino Acid Uptake and Metabolism of Legionella pneumophila Hosted by Acanthamoeba castellanii*

Background: Legionella replicates within vacuoles of host cells, but less is known about its metabolism during intracellular growth. Results: Isotopologue profiles of metabolites from Legionella grown in 13C-prelabeled amoebae provide fingerprints of intracellular metabolism. Conclusion: Host-derived amino acids are efficiently used for bacterial protein biosynthesis. Significance: Isotopologue profiling using 13C-prelabeled amoebae reflects key metabolic features of intracellular Legionella. Legionella pneumophila survives and replicates within a Legionella-containing vacuole (LCV) of amoebae and macrophages. Less is known about the carbon metabolism of the bacteria within the LCV. We have now analyzed the transfer and usage of amino acids from the natural host organism Acanthamoeba castellanii to Legionella pneumophila under in vivo (LCV) conditions. For this purpose, A. castellanii was 13C-labeled by incubation in buffer containing [U-13C6]glucose. Subsequently, these 13C-prelabeled amoebae were infected with L. pneumophila wild type or some mutants defective in putative key enzymes or regulators of carbon metabolism. 13C-Isotopologue compositions of amino acids from bacterial and amoebal proteins were then determined by mass spectrometry. In a comparative approach, the profiles documented the efficient uptake of Acanthamoeba amino acids into the LCV and further into L. pneumophila where they served as precursors for bacterial protein biosynthesis. More specifically, A. castellanii synthesized from exogenous [U-13C6]glucose unique isotopologue mixtures of several amino acids including Phe and Tyr, which were also observed in the same amino acids from LCV-grown L. pneumophila. Minor but significant differences were only detected in the isotopologue profiles of Ala, Asp, and Glu from the amoebal or bacterial protein fractions, respectively, indicating partial de novo synthesis of these amino acids by L. pneumophila. The similar isotopologue patterns in amino acids from L. pneumophila wild type and the mutants under study reflected the robustness of amino acid usage in the LCV of A. castellannii.

The Gram-negative bacterium Legionella pneumophila is an environmental aquatic pathogen that replicates within the cells of primitive host organisms such as Acanthamoeba castellanii, Hartmanella sp., Naegleria sp., or Dictyostelium discoideum (1)(2)(3). Nevertheless, Legionella is an opportunistic pathogen and inhalation of contaminated aerosols causes Legionnaires' disease, a severe pneumonia (4). Once in the lung macrophages, Legionella uses the same mechanisms required for replication as in amoebae (5). The bacteria inhibit the phagolysosomal pathway and build up a specialized compartment called Legionella-containing vacuole (LCV) 4 by secretion of dot/icm effector proteins. Inside the LCV, Legionella starts to replicate, and when nutrition becomes limiting, the bacteria switch to the transmissive form, which finally ends up in the so called mature intracellular form. The mature intracellular form exhibits a spore-like shape with a thick cell wall and high amounts of the cytoplasmatic energy and carbon storage poly-3-hydroxybutyrate (PHB) (6 -9).
In the replicative phase, the bacteria are metabolically highly active and resistant to sodium ions, a trait genetically related to avirulence (10). The maximum of metabolic activity is reached from mid to late exponential growth phase (11). In the transmissive phase, Legionella becomes metabolically dormant, sensitive against sodium ions, osmotically resistant, flagellated (motile), and cytotoxic. Subsequently, the virulent bacteria reach the host cytosol and are released from the host cell (12)(13)(14)(15). Transcriptional analyses reflect the biphasic life style of Legionella because half of the genes predicted in the genome showed a shift in gene regulation during the life cycle of the bacterium (16). During the replicative phase in amoebae, genes encoding proteins for the uptake and catabolism of peptides (amino peptidases and proteinases) are up-regulated, especially for the metabolism of Ser, Thr, Gly, Tyr, Ala, and His. Addition-ally, the genes encoding enzymes of the Entner-Doudoroff pathway and the glucoamylase GamA were found to be induced during replication in A. castellanii, indicating the usage of carbohydrate sources from amoebae (16). Indeed, GamA is responsible for glycogen-and starch-degrading activities of L. pneumophila and is expressed during replication in A. castellanii (17). The Legionella biphasic life cycle can be modeled in broth where the exponential phase corresponds to the replication of the bacteria in the LCV (13). Earlier studies indicated that Legionella uses amino acids as energy and carbon sources (11, 13, 18 -23). Under these in vitro conditions, Arg, Ile, Leu, Met, Thr, Ser, Val, and Cys were shown to be essential for L. pneumophila (11, 18 -25). Ser, Thr, Tyr, and particularly Glu were suggested as preferred carbon and energy sources (11,25). Indeed, external Cys, Gln, Ser, and Met, but also Val, Glu, Tyr, Arg, and Thr stimulated intracellular replication in the LCV of human macrophages (26,27). More recently, it was shown that the concentration of free amino acids in the cytosol of A. castellanii was increased because of infection with L. pneumophila activating the amoebal proteasome by the bacterial AnkB protein (28,29). Thus, the Legionella ankB null mutant had a severe defect for replication in Acanthamoeba, which was rescued by supplementation with various amino acids, pyruvate, or citrate. All of these results provide strong evidence for the requirement of host-derived amino acids for replication of Legionella in the LCV (26 -31). Indeed, some amino acids (Ile, Leu, Phe, Tyr, Val, Trp, Ser, Met, Gln and Cys) may enter the LCV, similar to in macrophages, through a putative SLC1A5-like transporter protein (26). Other amino acids may enter the LCV via yet unknown amino acid transporters (16,27,32,33). On the basis of the current experimental evidence, it was hypothesized that amino acids auxotrophy of L. pneumophila, and its host was synchronized during evolution (34,35). However, the transfer of host amino acids to intracellular L. pneumophila has not yet been directly observed, and the nature of the major nutrient(s) for L. pneumophila replication in the LCV of host cells is still obscure (for a recent review, see Ref. 36).
Under in vitro conditions, glucose was shown to also feed the central metabolism of L. pneumophila (11,23,24). The Entner-Doudoroff pathway appeared as the main route of glucose metabolism and turned out to be relevant for intracellular growth (23,37), but the genome of L. pneumophila also contains genes of an active glycolysis and pentose phosphate pathway (PPP) (23,38,39). Specifically, feeding of L. pneumophila with stable isotope ( 13 C)-labeled amino acids or glucose followed by determination of the resulting isotopologue pattern of key metabolites (e.g. amino acids and PHB) allowed identification of the rates and the pathways of serine and glucose usage under extracellular in vitro conditions (23). In the meantime, protocols have been established that enable isotopologue profiling for bacteria replicating under intracellular conditions, e.g. in the LCV of U937 macrophages or amoebae as host organisms (40).
Following these protocols, we have now used 13 C-prelabeled A. castellanii to investigate the nutrient usage and general carbon pathways of L. pneumophila Paris wild type and some of its mutant strains defective in putative key enzymes or regulators of carbon metabolism. With our study, the direct incorporation and usage of amino acids derived from the host is demonstrated for all strains under study of L. pneumophila growing in the LCV of A. castellanii.
Intracellular Replication/Survival Assay in A. castellanii-The intracellular multiplication/survival assay with and without competition was carried out as described previously (23).
DNA Techniques and Sequence Analysis-Genomic and plasmid DNAs were prepared according to standard protocols and the manufacturer's instructions (45). PCR was carried out using a TRIO-Thermoblock (Biometra, Göttingen, Germany) and Taq DNA polymerase (Qiagen). Foreign DNA was introduced into E. coli by electroporation with a gene pulser (Bio-Rad) according to manufacturer's specifications at 1.7 kV, 100 ⍀, and 25 mF. Both strands of plasmid DNA were sequenced with infrared, dye-labeled primers by using an automated DNA sequencer (LI-COR-DNA 4000; MWG-Biotech, Ebersberg, Germany). Primers were obtained from Eurofins MWG Operon (Ebersberg, Germany). The XbaI restriction enzyme was from New England Biolabs (Frankfurt, Germany).
Gene Cloning and L. pneumophila Paris ⌬keto Knock-out Mutant Construction-The knock-out mutants of gene lpp1788, annotated as a ␤-ketothiolase (EC 2.3.1.16), was con-structed as described before (16,46). In brief, lpp1788 (keto) was inactivated by insertion of a kanamycin resistance (kan R ) cassette into the chromosomal gene. The chromosomal region containing the lpp1788 gene and flanking regions (1149 and 1296 bp) were PCR-amplified with the primers Keto-fwd and Keto-rev. The product was cloned into the pGEM-T Easy vector (Promega) resulting in pIB-Keto1. On this template, an inverse PCR was performed using the primers Keto-inv-R2 and Keto-inv-U3 introducing an XbaI restriction site. The PCR product contained the pGEM-Teasy backbone and the chromosomal flanking regions of lpp1788. It was religated (pIB-Keto3) and XbaI-digested. A kan R cassette with XbaI restriction sites was cloned into pIB3 resulting in pIB-Keto4. For chromosomal recombination, the construct (i.e. PCR fragment containing the kan R cassette with flanking regions of the gene of interest) was amplified per PCR with the primers Keto-fwd and Keto-rev. Natural transformation of L. pneumophila Paris was done as described before with modification (47). In brief, 5 ml of an exponentially grown overnight culture was harvested by centrifugation at 1500 ϫ g for 10 min, and 3 ml of the supernatant was removed. The remaining culture was resuspended, transferred to a plastic tube, and incubated with the PCR product for 3 days at 30°C without agitation. Subsequently, bacteria were grown on antibiotic selective media for 4 more days at 37°C and 5% CO 2 . Screening for mutants obtained by homologous recombination was performed by PCR with kan R legionellae. Three independent ⌬keto mutant strains were generated and confirmed by PCR analysis with primer pairs binding outside of the recombinant site (Keto-outsideF/R) in combination with Keto-rev/fwd and KmU/R. All primers are listed in Table 1.
Isotopologue Profiling of L. pneumophila ⌬keto, ⌬rpoN, and ⌬fleQ in Medium Containing [U- 13 C 3 ]Serine and [U- 13 C 6 ]Glucose-The cultivation of all strains and the 13 C labeling experiments was performed according to Eylert et al. (23). Briefly, 1 ml of an overnight culture of the strains were cultured in 250 ml of AYE medium (23) supplemented with 2 g/liter of [U-13 C 6 ]glucose or 0.3 g/liter of [U-13 C 6 ]serine. Incubation was conducted at 37°C and 220 rpm. Bacteria were harvested after 29 h (stationary growth phase, see (23)), killed with sodium azide (10 mM), and pelleted at 5500 ϫ g at 4°C for 15 min. The pellet was washed twice with 200 ml of water and once again with 2 ml of water. The supernatant was discarded, and the bacterial pellet was autoclaved at 120°C for 20 min (23).
Isotopologue Profiling of L. pneumophila from 13 C-Prelabeled A. castellanii-A. castellanii was cultivated in 12-cell culture flasks (175 cm 2 ). A volume of 50 ml of PYG medium was inoculated (1:10) and incubated for 3 days at room temperature. The adhered amoebae were washed two times by gently shaking the flask containing 10 ml of Ac buffer. For prelabeling, the amoebae were incubated in Ac buffer supplemented with 11 mM [U-13 C 6 ]glucose for 24 h at 37°C. To remove external [U-13 C 6 ]glucose, the amoebae were washed two times with Ac buffer, and the flasks were incubated in 50 ml of fresh Ac buffer for 2 h at 37°C. L. pneumophila, grown on a BCYE agar plate for 3 days, was diluted in Ac buffer. The suspension was adjusted to an A 600 of 1, and 1 ml/flask was used for infection (about 10 9 cells/ml, resulting in an MOI of 100). After 1 h at 37°C, all extracellular bacteria were removed by washing two times with 10 ml Ac buffer. Subsequently, the infected amoebae were incubated for further 22 h at 37°C. The infection status was checked by light microscopy, and cells were harvested in late exponential phase of the bacteria just before lysis of the host cell. At this time point, an aliquot was plated out on LB agar plates, and the plates were incubated for 24 h to exclude the possibility of contaminations. Cells were transferred to 50-ml tubes and killed by addition of sodium azide at a final concentration of 10 mM. The amoebae were frozen for at least 2 h at Ϫ80°C and thawed in a 37°C water bath and vortexed for 20 s. Successful lysis was monitored microscopically, and intactness of the bacteria was checked via plating the suspension on BCYE agar plates without sodium azide treatment before and after the lysis step. Differential centrifugation allowed separation of unlysed amoebae and high density cellular components (fraction 1, F1), L. pneumophila (fraction 2, F2), and the cytosolic proteins of A. castellanii (fraction 3, F3) as follows. The suspension was centrifuged at 600 ϫ g and 4°C for 15 min. The pellet contained F1 and was washed three times with Ac buffer. The supernatant was transferred to a new tube and centrifuged at 3600 ϫ g and 4°C for 15 min. The resulting pellet (F2) was washed two times in distilled water, and the supernatant was filtrated through a 0.22-m pore filter to exclude contaminating bacteria. The supernatant was precipitated by addition of 100% TCA to a final concentration of 10% and incubated on ice for 30 min or overnight at 4°C. The denaturated proteins (F3) were spun down at 4600 ϫ g and 4°C for 30 min. All three pellets were autoclaved at 120°C for 20 min and stored at Ϫ20°C until Western blot fractionation control and work-up for GC-MS analysis. For Western blot fractionation control, the pellets generated from 12 flasks were solved in 600 l of distilled water, 10 l were used for SDS-PAGE, and the remaining solution was pelleted again for storage until GC-MS analysis.  A 600 ϭ 1) served as control (see Fig. 1A). To specify whether the contamination of the Legionella fraction (F2) was below 10%, which is indispensable to analyze samples in GC-MS, the noninfected Acanthamoeba cytosol (F3) was diluted in distilled water to imitate 0.5 and 5% contamination in a final volume of 10 l (see Fig. 1B). A. castellanii derived proteins were detected with polyclonal (1:200) (51) or a monoclonal antibody (␣-actin MAB1501 from Millipore; 1:1000). Legionella was detected using MONOFLUO anti-Legionella staining reagent from the Legionella IFA test kit (32.514; Bio-Rad), which contained a monoclonal antibody against the outer membrane protein Momp (1:100). The secondary HRP-linked antibodies were goat ␣-rabbit IgG (HϩL) for detection of A. castellanii and goat ␣-mouse IgG (HϩL) for detection of actin or Momp (Dianova, Hamburg, Germany) in a concentration of 1:5000. The blots were incubated with commercially available Pierce ECL Western blotting substrate ECL detection solution (Thermo Fisher Scientific) for 1 min and exposed to film. A Western blot analysis of the fractions was done before each GC-MS analysis to exclude possible cross-contaminations of the fractions.

SDS-PAGE and Immunoblotting Control Experiments-
Acid Hydrolysis and Derivatization for GC-MS Analysis-Pellet F2 or F3, respectively, was analyzed as described earlier (23). The resulting N-(tert-butyldimethylsilyl) amino acids and trimethylsilyl-3-hydroxybutyrate were then subjected to GC-MS analysis.
Mass Spectrometry-GC-MS analysis was performed with a GCMS-QP 2010 Plus (Shimadzu, Duisburg, Germany) as described earlier (23). Data were collected using the GCMS solution version 2 software (Shimadzu). Samples were analyzed at least three times. The theoretical isotope ratio and numerical deconvolution were computed according to standard procedures: (i) determination of the mass spectra of N-(tert-butyldimethylsilyl) amino acids and trimethylsilyl-3-hydroxybutyrate, respectively, (ii) determination of the mass isotopomer distribution of the labeled compounds, and (iii) correction for incorporation of 13 C from natural abundance (23,(52)(53)(54). Arg, Cys, Trp, and Met were not detected by using this method (54). Furthermore, acidic treatment used for protein hydrolysis converted Asn and Gln into Asp and Glu. The labeling data given for Asp and Glu therefore represent Asn/Asp and Gln/Glu averages, respectively.

Optimization of the 13 C Labeling and Infection Assay Using A. castellanii as Host
The amoeba A. castellanii is a natural host of Legionella in aquatic reservoirs. To analyze the transfer and usage of substrates from the host to intracellular L. pneumophila, we prelabeled A. castellanii with the stable 13 C-isotope by using [U-13 C 6 ]glucose as a supplement to the growth buffer. To receive high yields of 13 C-labeled amino acids in proteins of A. castellanii, we tested two different approaches. On the one hand, we grew A. castellanii for 3 days in PYG medium supplemented with 11 mM [U-13 C 6 ]glucose. Then the cells were split. This procedure was performed three times. On the other hand, A. castellanii cells were grown in PYG for 3 days, the medium was exchanged with Ac-buffer containing 11 mM [U-13 C 6 ]glucose, and the cells were further incubated for 24 h. Then the cells were harvested and protein-derived amino acids from both experiments were analyzed by GC-MS. 13 C-Excess values (mol %) are given in Table 2 (first and second columns). Because the second method afforded higher 13 C enrichments in amino acids, this setting was used to label A. castellanii in the experiments described below.
These prelabeled amoebae were then infected with L. pneumophila at a MOI of 100. After co-incubation for 2 h, remaining extracellular bacteria were removed by repeated exchange of the infection buffer. After 22 h of incubation in Ac buffer (without glucose), the amoebae were round shaped, and the LCV was filled with bacteria. Only 6 -8% of the A. castellanii cells harbored motile bacteria, indicating that the majority of the infected hosts were not yet lysed at the time of harvest.

Control of the Fractionation Protocol
The infected amoebae were harvested and lysed, and subsequently, the amoebal cytosol (F3) was separated from the still intact bacteria (F2) by differential centrifugation followed by sterile filtration (see "Experimental Procedures" and Ref. 40). The first centrifugation step ("low speed") resulted in fraction 1 (F1), which contained a mixture of high density compartments of the amoebal cytosol, amoebal ghosts, and unlysed amoebae with bacteria. Centrifugation of the supernatant ("high speed") resulted in pelleted bacteria (F2) and a further supernatant. This supernatant was sterile filtered to ensure the absence of bacteria, and then the dissolved proteins were precipitated from this fraction (resulting in F3).
To assess cross-contamination rates of the fractions, we performed Western blot analysis using a polyclonal ␣-A. castellanii antibody (␣-Ac), an A. castellanii specific monoclonal ␣-actin antibody and a L. pneumophila specific monoclonal ␣-major outer membrane protein antibody (␣-Momp). For a further control, the complete fractionation protocol was also performed with uninfected A. castellanii cells. An Acanthamoeba-specific protein pattern was detected in F1 and F3 using ␣-Ac antiserum, no matter whether the amoebae were infected or not. In the L. pneumophila fraction F2, only a minor contamination (caused by A. castellanii-specific proteins) was visible (Fig. 1, A and C). A cross-reaction of Legionella proteins was observed with the ␣-A. castellanii antiserum, but not with the actin antibody (Fig. 1A, F2 and Lp).
To verify the sensitivity of both ␣-Acanthamoeba antisera, we diluted the noninfected Acanthamoeba fraction F3 in H 2 O. We could determine that even a dilution of 0.5% of F3 was detectable in our Western blot by the ␣-A. castellanii and the ␣-actin antibody (Fig. 1B). At a concentration higher than 5%, the reaction reached its saturation point (Fig. 1D). The ␣-Momp antibody clearly detected Legionella in the fractions F1 and F2, but not in F3 (Fig. 1C, ␣-Momp). Because there was nearly no Acanthamoeba-specific signal detectable in fraction F2, the contamination of the Legionella fraction F2 seems to be below 0.5% and therefore applicable for isotopologue analysis.
In a further control experiment, we plated the bacteria before and after the amoeba lysis step and found no reduction in CFU, indicating that the bacteria were not lysed during this procedure. Altogether, we could therefore exclude a significant contamination of the sterile filtered Acanthamoeba cytosol fraction (F3) by Legionella-derived proteins (F2) and vice versa.

Metabolism and Amino Acid biosynthesis of A. castellanii
To determine the transfer of carbon substrates, e.g. amino acids from A. castellanii to the intracellular bacteria, by comparative isotopologue profiling, it was crucial to know the isotopic signatures of amino acids in the prelabeled host cells. Therefore, noninfected amoebae were labeled with [U-13 C 6 ]glucose following the same settings as used to prelabel amoebae for the infection experiments. Thus, A. castellanii were incubated for 24 h in Ac buffer containing 11 mM [U-13 C 6 ]glucose. The cells were then hydrolyzed under acidic conditions, and the resulting amino acids were analyzed by GC-MS. Of 14 detected amino acids, nine amino acids acquired significant 13 C label from [U-13 C 6 ]glucose indicating de novo biosynthesis of these amino acids, albeit at different rates ( Fig.  2A and supplemental Table S1). The 13 C excess values decreased from Ala (17%) Ͼ Glu (7.5%) Ͼ Asp (5%) ϭ Phe (5%) ϭ Pro (5%) ϭ Ser (5%) Ͼ Gly (4%) Ͼ Tyr (3%) Ͼ Thr (1%), whereas de novo synthesis from glucose was not detectable for His, Ile, Leu, Lys, and Val. When the amoebae were incubated for further 22 h in medium without [U-13 C 6 ]glucose (i.e. following the protocol of the infection assay), 13 C excess of Ala, Ser, and Gly was found reduced by a factor of about 2, whereas the values of the other labeled amino acids (Glu, Asp, Pro, Phe, Tyr, and Thr) remained nearly constant ( Table 2, third column). This shows that only Ala, Ser, and Gly were subjected to some metabolic turnover during the 22-h interval of the infection assay. The isotopologue compositions of the amoebal amino acids were important fingerprints to analyze the usage of these amino acids by the intracellular bacteria (see below), and therefore, these profiles are now described in some detail (see also Fig. 3A and supplemental Table S2a).
Alanine-The labeling pattern of Ala was characterized by a high abundance of the triple 13 C-labeled isotopologue (M ϩ 3). This species can be explained by uptake of [U-13 C 6 ]glucose by the amoebae, its conversion into the 6-phosphate, and degra- dation to [U- 13  Proline-The precursor for the de novo synthesis of Pro is ␣-ketoglutarate, and therefore the isotopologue composition reflects that of Glu.
Serine-The predominant M ϩ 1 labeling could indicate the action of Ser-hydroxymethyltransferase using unlabeled Gly and 13 C 1 -labeled methylene-tetrahydrofolate as substrates. The 13 C 3 -isotopologue was not abundant, which excludes major flux from [U-13 C 6 ]glucose via 13 C 3 -labeled glycolytic intermediates into Ser.
Glycine-Gly showed M ϩ 1 and M ϩ 2 isotopologues at similar relative amounts. This pattern can be explained by the interconversion of Ser and Gly, but the enrichment of the M ϩ 2 fraction suggests an alternative route for Gly biosynthesis from a 13 C 2 -labeled source.
Phenylalanine and Tyrosine-Phe and Tyr were synthesized by the shikimate pathway from PEP and erythrose 4-phosphate (E4P) via the intermediate chorismate. This is reflected by the isotopologue patterns of Phe and Tyr which are characterized by M ϩ 3 and M ϩ 4 isotopologues, respectively. The M ϩ 3 pattern is formed when [ 13 C 3 ]PEP and unmarked E4P enter the pathway, whereas M ϩ 4 labeling occurs when [ 13 C 4 ]E4P, derived from [ 13 C 6 ]fructose 6-phosphate, is condensed with unlabeled PEP. Thus, the data demonstrate an active shikimate pathway in the eukaryotic amoeba, but also efficient flux via the transketolase reaction in the PPP to provide the E4P unit. Generally, infection with L. pneumophila wild type or its isogenic mutants ⌬zwf, ⌬keto, ⌬fleQ, or ⌬rpoN did not significantly change the labeling patterns of amino acids from the host, as described above (Figs. 2A and 3A and supplemental Tables S1 and S2).

Amino Acid and PHB Metabolism of Intracellular L. pneumophila
The metabolism of intracellular Legionella was investigated with L. pneumophila Paris and its isogenic ⌬fleQ, ⌬rpoN, ⌬zwf, and ⌬lpp1788 (⌬keto) mutant strains (43,55). L. pneumophila Paris was selected because various micro array data of in vitro and in vivo (A. castellanii) grown bacteria were already published for this strain (16,32,43). The mutant strains were selected because the deleted genes refer to putative key regulators or enzymes of carbon metabolism. Thus, FleQ and RpoN are regulatory proteins of the flagellar regulon (43,56), which were also shown to be involved in the expression of amino acid transport proteins and genes of the His and Arg biosynthesis pathways (43). Furthermore, ⌬fleQ and ⌬rpoN mutant strains of L. pneumophila were able to replicate in a general infection assay within A. castellanii, but both mutants showed a reduced fitness in competition to the wild type strain (55). The ⌬zwf gene encodes the first enzyme of the Entner-Doudoroff pathway involved in glucose metabolism and was also shown to exhibit a reduced fitness in competition with the wild type strain (23).
In the ⌬keto strain, the gene encoding a ␤-ketothiolase (lpp1788) was inactivated (this study). This ␤-ketothiolase was proposed to be involved in fatty acid degradation and PHB bio-synthesis/degradation. It was not yet known whether this gene influences the intracellular replication of L. pneumophila. Therefore, we analyzed the ⌬keto strain in the intracellular replication/survival assay with and without competition. We could demonstrate that the mutant behaved similar as the wild type strain in these assays (Fig. 4). Thus, all L. pneumophila strains analyzed in this study grew similar in the general infection assay. With the exception of the ⌬keto and ⌬gamA strains, all other mutant strains investigated in this study showed a reduced fitness when growing in competition with the wild type strain (Fig. 4) (55).
Because the impact of the ⌬keto, ⌬rpoN, and ⌬fleQ mutations on the metabolic network of L. pneumophila was not known at all, we first analyzed these strains under in vitro conditions by isotopologue profiling using [U-13 C 3 ]serine or [U-13 C 6 ]glucose as supplements to the growth medium (23). In the experiment with [U-13 C 3 ]serine, the labeling patterns of amino acids from the ⌬keto, ⌬rpoN, and ⌬fleQ mutants were apparently identical with the respective patterns of amino acids from an experiment with the wild type strain (Fig. 5A and sup- plemental Tables S3 and S4). However, using [U-13 C 6 ]glucose as a substrate, the 13 C excess values of Ala, Asp, and Glu from the ⌬keto, ⌬rpoN, and ⌬fleQ mutant strains were higher as compared with the wild type strain (Fig. 5B, Table 3, and supplemental Tables S3 and S4). For PHB, the major difference was that in the ⌬keto strain, the 13 C excess value was only 28 and 25% of that of the wild type strain, using [U-13 C 3 ]serine or [U-13 C 6 ]glucose as substrates, respectively (Table 3). In contrast, the ⌬rpoN and ⌬fleQ mutants displayed higher 13 C enrichments (160 and 143%, respectively) in PHB from the experiment with [U-13 C 6 ]glucose in comparison with the cor-responding experiment with the wild type (Fig. 5B, Table 3, and  supplemental Tables S3 and S4).
In the next step, we determined and compared the isotopologue profiles of amino acids in the amoebal and bacterial protein fractions (F3 and F2, respectively) from the in vivo infection experiments using 13 C-prelabeled amoebae as hosts. Generally, the 13 C isotopologue patterns of amino acids from the analyzed L. pneumophila strains (F2 fractions, wild type, and mutants) resembled the respective patterns in amino acids from A. castellanii (F3 fractions) from the same experiment (Figs. 2 and 3). Thus, unlabeled amino acids in the host were found unlabeled also in the bacteria, whereas labeled host amino acids were 13 Cenriched in the protein fraction of L. pneumophila. Both the molar abundances and the isotopomer compositions of the labeled amino acids from amoebae or bacteria were similar if not identical, respectively, with the exception of Ala, Asp, and Glu. To better visualize these differences, we calculated the ratios of 13 C excess (mol %, mean value of all experiments) in amino acids from bacterial and amoebal proteins (F2/F3). As shown in Table 2, these ratios were 0.57 (Asp), 0.60 (Glu), 0.87 (Ala), 0.95 (Ser, Gly), 1.06 (Phe), and 1.09 (Pro). Thus, in particular the Legionella-derived amino acids Asp and Glu showed a significant lower 13 C excess value than in A. castellanii (Fig. 6). However, not only the 13 C excess values were different. Whereas the relative amounts of M ϩ 1 in Asp and Glu from F2 and F3 were similar, the M ϩ 2 and M ϩ 3 fractions were significantly lower in the bacterial fraction F2 (Fig. 3 and supplemental Tables S1 and S2). Together with our careful fractionation control by Western blotting, the differences in Asp and Glu labeling confirmed that the detected similar labeling patterns of other amino acids from the bacterial and the host cell fraction were not due to cross-contaminations. However, more importantly, the decrease in 13 C incorporation and the changes in the isotopologue profiles indicate that these amino acids were not only taken from the host and ready-made incorporated into bacterial protein, but that especially Asp and Glu were also made de novo by the intracellular bacteria from unlabeled compounds and metabolic intermediates provided from the host cell. The increase in the relative amounts of M ϩ 1 isotopologues probably reflects contributions of 13 C 1 -labeled precursors from the TCA of L. pneumophila. It can be concluded that either these intermediates or related metabolites are transferred from the host cell or that the labeled host amino acids were catabolized by L. pneumophila providing 13 C 1 -labeled precursors for Asp and Glu biosynthesis. Another interesting finding is that for the amino acids Phe and Pro, which in vitro were rarely synthesized using [U-13 C 3 ]serine or [U-13 C 6 ]glucose as substrates (Fig. 5), in vivo the 13 C excess values were significantly higher in both fractions (Fig. 2). Again, this indicated the usage of host cell made Phe and Pro in the synthesis of bacterial proteins under in vivo conditions.
Whereas the 13 C excess values of Phe and Pro seemed to be slightly reduced in the ⌬keto strain but increased in the ⌬fleQ and ⌬rpoN mutants (Fig. 2), the profiles of the other bacterial amino acids were almost unchanged in comparison with the respective patterns of the wild type strain. With the ⌬zwf strain, we could not detect any differences in the 13 C profiles notably in sharp contrast to the earlier in vitro results (23).
The 13 C excess values of PHB (measured as a derivate of 3-hydroxybutyrate from the acidic hydrolysate of F2) were only about 1 mol % with a dominating M ϩ 1 and M ϩ 2 pattern (ratio 1:1) in the wild type, ⌬zwf, and ⌬rpoN strains (supplemental Tables S1 and S2). Nevertheless, these enrichments reflected synthesis of bacterial PHB from host-derived 13 C-labeled substrates, obviously via 13 C 2 -and 13 C 1 -acetyl-CoA. The 13 C enrichments of PHB in the ⌬keto and ⌬fleQ strain slightly differed and were ϳ0.7 and 1.5%, respectively (supplemental Table S1).

DISCUSSION
Metabolism of A. castellanii-Acanthamoeba is a chemoheterotroph organism and metabolizes glucose mainly via glycolysis and the PPP. Pyruvate can be converted into acetyl-CoA, which enters the TCA cycle. Using these pathways, the carbon backbone of many amino acids generated by Acanthamoeba are provided (57,58). Indeed, isotopologue profiling performed in this study showed that label from exogenous [U-13 C 6 ]glucose is transferred to Ala, Ser, Glu, Asp, Pro, Gly, Phe, Tyr, and Thr, whose precursors stem from degradation products of glucose via glycolysis, the PPP, and the TCA pathway, respectively. In the non-nutrient Ac buffer, [U-13 C 6 ]glucose was more efficiently used by A. castellanii as a precursor for amino acids than in the nutrient-rich PYG medium. A further incubation of 13 Clabeled A. castellanii in unlabeled Ac buffer (i.e. without any glucose) caused a decrease in 13 C-excess ( Table 2, first through third columns) predominantly in Ala, Gly, and Ser, probably because of a further de novo synthesis of these amino acids from the nonlabeled citrate present in the Ac buffer (1 g/liter). We could not detect de novo synthesis of Ile, Leu, Val, His, and Lys from [U-13 C 6 ]glucose in A. castellanii. Obviously, the amoeba is auxotrophic for these amino acids. Indeed, it was already shown earlier that Acanthamoeba minimal medium needs to be supplemented with Ile, Leu, Val, Arg, and Met, as long as glucose is present as a single carbon source (59). If acetate is the carbon source, then also Gly was needed for growth in minimal medium. Notably, we could not detect 13 C-labeled His, although the genes for His biosynthesis are present within the genome sequence of A. castellanii ATCC30010 (60, 61) that was published recently. On the other hand, genes for the shikimate pathway are present in the genome (61), and we could indeed detect labeled Phe and Tyr in the experiment with [U-13 C 6 ]glucose. To our knowledge, this is the first direct experimental demonstration of de novo synthesis of aromatic amino acids in A. castellanii. In accordance, it has been observed earlier that some pathogenic protozoa including A. castellanii show reduced growth by the presence of inhibitors of the shikimate pathway (62)(63)(64). Interestingly, Naegleria gruberi, another natural host for Legionella, is negative for the shikimate pathway as shown by genome sequence analysis (65).
On the basis of the similar labeling profiles in experiments with uninfected or infected A. castellannii, the amino acid and core metabolism (providing precursors of the nonessential amino acids) of Acanthamoeba was not significantly modulated by intracellular (i.e. in the LCV) replicating Legionella. This is in contrast to earlier observations with Chlamydia pneumoniae, Listeria monocytogenes, or Salmonella enterica serovar typhi- FIGURE 5. A, labeled isotopologues in amino acids and PHB of L. pneumophila Paris WT grown in AYE medium or the isogenic mutant strains Lp ⌬keto, ⌬fleQ, and ⌬rpoN labeled in vitro with [U-13 C 3 ]serine. B, labeled isotopologues in amino acids and PHB of L. pneumophila Paris WT grown in AYE medium or the isogenic mutant strains Lp ⌬keto, ⌬fleQ, and ⌬rpoN labeled in vitro with [U-13 C 6 ]glucose. Overall 13 C excess (mol %) of labeled isotopologues amino acids and PHB measured by GC-MS analysis. The color map indicates 13 C excess in quasilogarithmic form to show even relatively small 13 C excess values. Each sample (from two different labeling experiments) was measured three times; the color for each amino acid correlates with the mean value of the three measurements. Amino acids represented by white boxes containing a numerical value; the numerical value represents the overall 13 C excess (mol %). Amino acids represented by white boxes could not be measured. murium, which modulated carbon fluxes of host metabolism, in particular fluxes via glycolysis and glutaminolysis of primary cells (66,67).
Metabolism of L. pneumophila in the LCV of 13 C-Prelabeled A. castellanii-To determine the uptake and usage of host-derived amino acids, 13 C-prelabeled A. castellanii were infected by L. pneumophila wild type strain or mutant strains and incubated for 22 h in medium without 13 C tracers. This setting determines that the majority of LCVs is not lysed and L. pneumophila at the end of the nonmotile replicative phase with high metabolic activity (14).
In our infection experiments, we estimated the growth phase of L. pneumophila by determining the motility of the bacteria within the amoeba at the time of harvest, indicating that not more than 8% of the analyzed bacteria had entered the transmissive phase. It is known that L. pneumophila becomes flagellated when entering the transmissive phase (12,15), and it has been published recently that this happens in vivo when the bacteria enter the cytosol (14).
Initial MS analysis revealed a very similar pattern of labeled amino acids from intracellularly grown L. pneumophila strains and their host cells. Therefore, we analyzed in considerable detail whether these results were due to severe cross-contamination of the analyzed fractions. The Legionella-specific ␣-Momp antibody clearly detected Legionella in fractions F1 and F2, as expected, but not in the A. castellanii fraction F3. In addition, no PHB could be detected in fraction F3. Furthermore, there was no specific signal visible in F2 using the Acanthamoeba-specific ␣-actin antiserum. The purity of the fractions was further confirmed by the finding that three L. pneumophila-derived amino acids (Ala, Asp, and Glu) showed significantly different 13 C excess values than the respective A. castellanii-derived amino acids (Fig. 6).
For a discussion of the observed labeling patterns, the reader is referred to Fig. 7. In prelabeled A. castellanii cells, we detected significant 13 C enrichments and specific isotopologue profiles (Fig. 7, asterisk) in Ala, Asp, Glu, Ser, Phe, Tyr, Pro, Gly, and Thr. On the other hand, Val, Leu, Ile, Lys, and His did not acquire any 13 C label. These labeling patterns and fingerprints provided the basis to study the transfer of amino acids as nutrients for intracellular Legionella. The detection of the same unlabeled and labeled amino acids with their specific isotopologue fingerprints in amino acids from the bacterial fraction indicated that the whole set of amino acids were transferred Our analysis demonstrated that the marked amino acids (*) are 13 C-labeled during the incubation of A. castellani in Ac buffer containing [U-13 C 6 ]glucose. As shown recently, the proteasome is involved in the generation of free amino acids in infected host cells (28). These amino acids (red circles) then could reach the LCV (Ac-LCV) by a SLC1A5-like amino acid transport protein (26) or by other yet not known amino acids transporters (black circles). For L. pneumophila (Lpp), many amino acid transport proteins (green circles) are known to be expressed during intracellular growth (16,33). Amino acids of intracellular grown L. pneumophila are given in black, and results of AYE grown bacteria are given in blue (from Ref. 23). Amino acids in parentheses could not be analyzed by the method used in this study. Phe and Tyr are underlined, indicating that these amino acids were unlabeled in vitro but labeled during in vivo growth of L. pneumophila. A few L. pneumophila-derived amino acids (#) showed a significant different isotopologue profile when compared with L. pneumophila-infected A. castellanii-derived amino acids. This seems to be a result of the metabolism of these amino acids by L. pneumophila through the Glu (step 1) and the Ala (step 2) transaminases or by de novo synthesis. PHB was also found to be labeled (yellow circle), presumably using Ac-CoA as the substrate, as demonstrated recently for in vitro growth (23). We do not have experimental data that support or rebut metabolization of carbohydrates, glycerol, or fatty acids during intracellular replication, but sugar, glycerol-3-phosphate, and fatty acid transport proteins are known for L. pneumophila (shown on the left), and glucose is metabolized during growth in vitro (23). To our knowledge, nothing is known about nutrients delivered by the cargo of endoplasmic reticulum vesicles to the LCV. Genes (Lpp gene numbers) encoding transporter proteins are given within the blue and green circles. AA, amino acids; Ac, A. castellanii; Lpp, L. pneumophila Paris; Hex-P-Tp, hexosephosphate transport protein; ER, endoplasmic reticulum; FA, fatty acids; KG, ketoglutarate; Pyr, pyruvate; CIT, citrate; MAL, malate; 1, Glu-Asp transaminase; 2, Glu-Pyr transaminase.
from the host cytosol to the LCV, taken up by L. pneumophila and subsequently directly used for protein biosynthesis. This close correlation of the labeling patterns in amino acids from A. castellannii and its intracellular L. pneumophila underlines the hypothesis of a synchronized evolution of amino acid metabolism including auxotrophy in intracellular bacteria and their host organisms (34,35). However, the differences in the isotopologue profiles of Asp and Glu suggested that a fraction of these amino acids was made de novo from other host carbon sources, which are not 13 C-enriched or are less 13 C-enriched, in particular from metabolites feeding the TCA and contributing one 13 C atom (Fig. 7). In a potential scenario, Asp and Glu are required for downstream biosynthetic pathways and must therefore be refilled by own bacterial biosynthesis. This is reasonable because it was postulated that the glutamate-aspartate transaminase reaction is associated with the TCA, anabolic, and gluconeogenetic pathways (39,68). Thus, the respective enzyme is highly active in medium grown L. pneumophila strains (68) and Glu plays a key metabolic role in vitro, is rapidly taken up, metabolized, and distributed in various cellular fractions within Legionella. Furthermore, Glu is the only amino acid that permits growth when added to chemically defined medium containing just eight essential amino acids (12,24,68). These observations illustrate that Glu is a key metabolic amino acid for the lifestyle of Legionella. We did preliminary in vitro experiments using 15 N-Glu revealing transfer of label into Ala and Asp, respectively (data not shown). Indeed, this suggests activity of the Glu/Pyr and/or Glu/Asp transaminases (Fig. 7). Nevertheless, Wieland et al. (26) found that the addition of Glu to L. pneumophila infected human macrophages has only a minor supporting effect on intracellular replication of Legionella. Cys, Gln, Ser, Met, and Arg were better inductors of bacterial replication. Interestingly, the enzymes of the Asp biosynthesis pathway were induced during growth within the LCV of human macrophages, supporting our hypothesis that Asp is synthesized by Legionella during in vivo growth (33). Our model is also supported by the fact that, under in vitro conditions, L. pneumophila is auxotrophic for Phe and Tyr (23), and in A. castellanii-grown L. pneumophila strains, both amino acids exhibited the same 13 C labeling pattern as the respective A. castellanii-derived amino acids. Thus, we could demonstrate for the first time a direct usage of many A. castellanii-derived amino acids for protein biosynthesis in L. pneumophila, corroborating other findings suggesting that host-derived amino acids are taken up by L. pneumophila and are important for the growth of the bacteria within the LCV of the host cell (16,26,28,33). Indeed, the induction of amino acid transporter genes was observed for several pathogens. Therefore, the in vivo recruitment of host amino acids seems to be a widely used strategy during intracellular replication (69 -73). As prominent examples, the usage of amino acids from the host has been shown for L. monocytogenes and Francisella tularensis, both replicating within the cytosol of the host cell (73)(74)(75).
We also analyzed four different mutant strains of L. pneumophila Paris. The inactivated genes encode regulators (fleQ and rpoN) or enzymes (zwf and keto) of different metabolic pathways. RpoN is an alternative factor ( 54 ) that needs an activator protein, like FleQ, FleR, or PilR for transcription initiation (56,76). Our recently published transcriptome data of in vitro grown L. pneumophila Paris strains revealed that RpoN influences the expression of genes relevant for arginine biosynthesis, glycolysis, gluconeogenesis, and the PHB synthesis pathway (43,77). Additionally, both regulatory proteins are necessary for full in vivo fitness in A. castellanii (55,56). The ⌬keto strain lacks lpp1788, which is one of two homologous ␤-ketothiolase genes (lpp1555 and lpp1788), putatively catalyzing the acetylation of acetyl-CoA. Our in vitro experiments showed for the ⌬keto strain a slightly delayed growth in AYE medium but not in intracellular growth within A. castellanii (Fig. 4). However, despite these differences in growth behavior, the amino acid labeling pattern of all examined Legionella strains grown in the LCV of 13 C-prelabeled A. castellannii showed comparable results. Interestingly, also the 13 C excess and labeling patterns of amino acids analyzed from the ⌬zwf strain grown within A. castellanii did not change from that of the wild type strain, whereas we recently showed significant differences under in vitro conditions using [U-13 C 6 ]glucose as a tracer (23). In vitro, L. pneumophila is able to metabolize glucose via the Entner-Doudoroff pathway. However, during the replicative phase within the LCV, L. pneumophila does not seem to use external glucose, because the reduced labeling of amino acids in the ⌬zwf mutant strain observed in vitro was not visible in vivo when [U-13 C 6 ]glucose was added to the Ac buffer during infection (preliminary results; data not shown). This is in good agreement with earlier results (37) and our own data (preliminary results; data not shown), indicating that L. pneumophila metabolize glucose predominantly in the nonreplicative transmissive phase of growth. Thus, it can be speculated whether glucose or related carbohydrates are used during the transmissive phase in the cytosolic compartment of the host cells.
PHB is the universal energy and carbon storage compound of Legionella and could be detected in the dichloromethane extracts of the Legionella fraction (F2) via 13 C NMR spectroscopy or MS spectrometry of 3-hydroxybutyrate in the hydrolysate (7,23,78). In vitro labeling experiments of Legionella with [U-13 C 3 ]serine and [U-13 C 6 ]glucose confirmed the known PHB formation from acetyl-CoA (23). Furthermore, PHB can be detected within Legionella in varying amounts from replicative to transmissive phase depending on the carbon source (2, 16, 78 -80). When using glucose or Ser as 13 C-labeled substrate, the 13 C excess value of PHB of the ⌬keto mutant strain was only 25-28% of that of the wild type strain. On the other hand, using glucose as the substrate, the 13 C excess value of the ⌬fleQ and ⌬rpoN mutant strains was 140 -160% of that of the wild type strain. Although the 13 C excess values of PHB in vivo were low, the 13 C excess values of the ⌬keto mutant strains were comparable with that of the in vitro experiments. The increased PHB labeling in the ⌬rpoN mutant in vitro may be explained by transcriptome data showing that the regulatory proteins FleR and RpoN repress the expression of genes putatively encoding acetoacetyl-CoA reductases (lpp0621, lpp2322, and lpp0620) and therefore PHB synthesis (43). The determined low in vivo 13 C incorporation rate into PHB of wild type Legionella is in accordance to the growth phase, because in our experiments the bacteria were harvested before they reached the cytosol, and this time interval corresponds to the replicative growth phase where PHB synthesis is still low.
As demonstrated for the analyzed amino acids and in contrast to in medium grown bacteria (23), there was no difference in the 13 C excess values of PHB from in A. castellanii grown ⌬zwf mutant strain and wild type (23). This result again corroborates the major usage of host-derived amino acids but not of glucose by Legionella during the replicative phase in the LCV.
Conclusion-Numerous studies about the usage of amino acids as carbon and energy source have been carried out in vitro and to a lesser extend in vivo, but the detection was always of indirect nature or based on transcriptome data. In our study, we have used 13 C isotopologue profiling, which directly reflects metabolic pathways and fluxes. On the basis of these data, we show that Legionella recruits a large set of amino acids including Ala, Asp, Glu, Ser, Phe, Tyr, Pro, and Gly from the Acanthamoeba cytosol surrounding the LCV and incorporates these amino acids directly into bacterial protein. In addition, the data indicate that Ala, Asp, and Glu are also synthesized or co-metabolized by the bacteria from other host-derived carbon bodies. These results illustrate special features of intracellular replication. Because Legionella naturally is not able to replicate outside of the LCV, nutrition supply inside of the specialized vacuole is one of the fundamental aspects of the host-pathogen interaction. Pathogenesis research progressively focuses on the formerly neglected topic of metabolism, because colonization of host cell tissues requires the adaptation of bacterial metabolism. Knowledge about the bacterial metabolic routes used in vivo could identify essential enzymes for the bacterial metabolism and, therefore, promising targets for antibiotic development.