Type II Secretion-Dependent Aminopeptidase LapA and Acyltransferase PlaC Are Redundant for Nutrient Acquisition during Legionella pneumophila Intracellular Infection of Amoebas

ABSTRACT Legionella pneumophila genes encoding LapA, LapB, and PlaC were identified as the most highly upregulated type II secretion (T2S) genes during infection of Acanthamoeba castellanii, although these genes had been considered dispensable on the basis of the behavior of mutants lacking either lapA and lapB or plaC. A plaC mutant showed even higher levels of lapA and lapB transcripts, and a lapA lapB mutant showed heightening of plaC mRNA levels, suggesting that the role of the LapA/B aminopeptidase is compensatory with respect to that of the PlaC acyltransferase. Hence, we made double mutants and found that lapA plaC mutants have an ~50-fold defect during infection of A. castellanii. These data revealed, for the first time, the importance of LapA in any sort of infection; thus, we purified LapA and defined its crystal structure, activation by another T2S-dependent protease (ProA), and broad substrate specificity. When the amoebal infection medium was supplemented with amino acids, the defect of the lapA plaC mutant was reversed, implying that LapA generates amino acids for nutrition. Since the LapA and PlaC data did not fully explain the role of T2S in infection, we identified, via proteomic analysis, a novel secreted protein (NttD) that promotes infection of A. castellanii. A lapA plaC nttD mutant displayed an even greater (100-fold) defect, demonstrating that the LapA, PlaC, and NttD data explain, to a significant degree, the importance of T2S. LapA-, PlaC-, and NttD-like proteins had distinct distribution patterns within and outside the Legionella genus. LapA was notable for having as its closest homologue an A. castellanii protein.

When we examined, in a similar way, gene expression in a plaC mutant (NU420) (23), we detected an even higher level of expression of lapA and lapB, while the expression levels of other T2S genes were unchanged or slightly decreased ( Fig. 2A). In contrast, when we assayed transcript levels in a lapA lapB mutant (NU324) (30), there was a further heightening in levels of plaC mRNA (Fig. 2B). We also detected an increase in celA transcription in the lapA lapB mutant (Fig. 2B), but, when we examined expression in a celA mutant (NU353) (29), there was no compensatory increase in expression of lapA or lapB or of any other genes (Fig. 2C). These data suggested that the intracellular role(s) of LapA/LapB might be functionally compensatory with respect to that of PlaC, such that when the aminopeptidases are absent there is an increase in acyltransferase levels and vice versa. When the plaC mutant and lapA lapB mutant were grown in broth, no compensatory changes in expression were seen (Fig. 2D), implying that the functional redundancy between PlaC and LapA/LapB may be specific to intracellular growth.
Double mutants reveal key roles for LapA and PlaC in infection of A. castellanii. We previously observed that neither plaC mutant NU420 nor lapA lapB mutant NU324  is impaired for infection of A. castellanii (23,30). Given our transcriptional data, we posited that a lapA plaC mutant, a lapB plaC mutant, and/or a lapA lapB plaC mutant might be impaired for infection due to the combined loss of two or more functionally redundant genes. Thus, we made double and triple mutants, and after observing their capacity to grow typically in BYE medium and survive normally in conditioned amoebal medium (see Fig. S1A and B [left panels] in the supplemental material), we tested them for infection of A. castellanii. Whereas the lapB plaC mutant (NU433) grew similarly to the wild type (Fig. 3A), the lapA plaC mutant (NU435) displayed an~50-fold defect in infection (Fig. 3B). The lapA lapB plaC mutant (NU436) was no more defective than the lapA plaC mutant (Fig. 3B). Since lapA and plaC are monocistronic ( Fig. S2A and B), the mutant defect was likely due to loss of lapA and plaC rather than polarity on downstream genes. When lapA was reintroduced into the lapA plaC mutant, there was restoration of infection ( Fig. 3C), confirming the role of LapA. Similarly, the lapA plaC mutant was complemented by reintroduction of plaC (Fig. 3D), affirming the role of PlaC. Thus, we extended our transcriptional analysis of intracellular legionellae to include N. lovaniensis (Fig. 1C). Again, we saw that lapA, lapB, and plaC were upregulated (Fig. 1C, right). legP and map were also notably upregulated (Fig. 1C, right). Previously, we determined that the lapA lapB mutant is not impaired for infection of N. lovaniensis, whereas the plaC mutant has an~10-fold defect (23). Given the transcription data, we compared the lapA lapB plaC mutant and plaC mutant for infection of N. lovaniensis, to possibly uncover another role for LapA (or LapB). However, the triple mutant was no more impaired than the single mutant (Fig. S3A). A like result was attained when we tested mutants in V. vermiformis (Fig. S3B). When analysis was expanded to include infected macrophages (Fig. 1D), lapA and lapB appeared to be among the most highly upregulated genes (right), whereas the change seen with plaC was modest (middle). Since the lapA lapB mutant and plaC mutant are not impaired in U937 cells (30,34), we compared the lapA lapB plaC mutant to the wild type with respect to infection of macrophages. However, no new mutant phenotype was detected (Fig. S3C). Thus, though lapA and plaC expression was upregulated upon Legionella growth in two hosts, the compensatory link between LapA and PlaC was detected only in A. castellanii. Our mutant analysis, which was based on transcript data, uncovered a role for PlaC in infection of A. castellanii, expanding the idea of the protein's importance beyond its role in infection of V. vermiformis and N. lovaniensis (23). A glycerophospholipid-cholesterol and -ergosterol acyl-transferase that has phospholipase A and lysophospholipase A activities, PlaC has been well studied biochemically (34,35). More importantly, our transcriptional and mutational analyses revealed, for the first time, a role for the LapA aminopeptidase in an intracellular infection event; thus, we focused our next efforts on defining the biochemical, structural, and phylogenetic traits of LapA.
LapA is a protease-activated, broad-spectrum aminopeptidase. Past work found that LapA is active against leucine, phenylalanine, and tyrosine aminopeptides, based on the ability of wild-type supernatants but not lapA mutant supernatants to cleave those substrates (30). LapA is annotated as a leucine aminopeptidase, and the Merops database assigns LapA as a member of the M28 family of peptidases (subfamily E) (28,30,36). We more completely defined the specificity of LapA by testing wild-type, lspF mutant, and lapA mutant supernatants against additional aminopeptides. Wild-type samples had activity against alanine, aspartate, glycine, isoleucine, lysine, methionine, serine, and valine aminopeptides, in addition to leucine and phenylalanine aminopeptides (Fig. 4A). They had no activity against proline-or glutamate-containing substrates. The lspF mutant lost activity against aspartate, isoleucine, leucine, lysine, methionine, phenylalanine, and valine substrates but showed normal activity against alanine, glycine, and serine aminopeptides (Fig. 4A). Thus, 7/10 of the activities are mediated by T2S. The lapA mutant lacked activity against aspartate, isoleucine, methionine, and valine aminopeptides and cleaved alanine, glycine, lysine, and serine substrates (Fig. 4A). As described before (30), the mutant lacked activity against leucine and phenylalanine aminopeptides (Fig. 4A). When the complemented lapA mutant was examined, we saw restoration of the six activities that were lacking in the mutant (Fig. 4A). The complement trended toward elevated levels of activity against alanine and serine aminopeptides (Fig. 4A), suggesting that increased amounts of LapA might cleave additional substrates.
To confirm the activities of LapA, we purified recombinant LapA-His6-tagged protein containing 378 amino acids of the mature lapA product (minus the 23-residue signal sequence). As expected, purified LapA had activity against leucine, valine, isoleucine, aspartate, methionine, and phenylalanine aminopeptides, with its activity being greatest against the leucine substrate (Fig. 4B). LapA lacked activity against proline, glutamate, and glycine, compatible with supernatant tests. Although assaying of supernatants had not revealed LapA-dependent cleavage of alanine, lysine, and serine aminopeptides, due to the presence of LapB and a T2S-independent factor (Fig. 4A) (30), the testing of purified enzyme affirmed LapA's ability to cleave these substrates (Fig. 4B). Combining these data with past results (30), we conclude that LapA is active against Ն10 substrates. Most substrates had nonpolar, amino acid targets, but, given its cleavage of aspartate, lysine, serine, and tyrosine aminopeptides, LapA has rather broad activity (37,38). Typically of a member of the M28 family (39), LapA was inhibited by bestatin (Fig. S4A).
LapA was predicted to contain an N-terminal, "protease-associated" (PA) domain, which in some M28 peptidases is necessarily excised in order to release a C-terminal peptidase domain (40,41). That LapA might be processed is suggested from our prior two-dimensional (2-D) PAGE analysis of supernatants, which had detected truncated forms of LapA (28). To determine if processed LapA has activity, we purified a version of LapA that was missing the putative PA domain and repeated the enzyme assays. Truncated LapA had activity against the same substrates that had been cleaved by full-length LapA, with leucine aminopeptide remaining as the favored substrate (Fig. 4C). Interestingly, the activity against methionine aminopeptide and, to a lesser extent, the phenylalanine aminopeptide was elevated, whereas the activities against the other seven substrates were similar or slightly decreased ( Fig. 4B and C). These data indicate that the removal of the PA domain of LapA, though not required, yields an enzymatically active protein. Also, full-length LapA and truncated LapA may have different substrate preferences.
We posited that the T2SS-dependent ProA protease might be cleaving secreted LapA into its N-terminal PA domain and C-terminal peptidase domain. ProA is the major secreted protein of L. pneumophila and is known to cleave other exoproteins (19,35,42,FIG 4 Aminopeptidase activities from L. pneumophila culture supernatants and purified recombinant LapA. Filter-sterilized supernatants from late-exponentialphase broth cultures of wild-type 130b (WT), lapA mutant NU320 (lapA), complemented lapA mutant NU320 containing plasmid-carried lapA (lapA/lapAϩ), and lspF mutant NU275 (lspF) were assayed for activity directed toward aminopeptidase substrates, with the N-terminal amino acid (AA) residue given. (A) Changes in background-adjusted absorbance at OD 405 for the AA-p-nitroanilide substrates or fluorescence (expressed in relative fluorescence units [RFU]) for the AA-␤-naphthylamide substrates were monitored kinetically over 2 h. Data are presented as means with standard errors of results from three independent experiments (n ϭ 3 each). Asterisks indicate significant reductions in activity of a mutant relative to the WT (*, P Ͻ 0.05; **, P Ͻ 0.01; ***, P Ͻ 0.001 [Student's t test]). (B and C) Strains with purified, full-length LapA (FL-LapA) (B) or LapA missing its N-terminal PA domain (ΔN-LapA) (C) were incubated for 2 h at 37°C with various AA-p-nitroanilide substrates (left panels) or AA-␤-naphthylamide substrates (right panels), and the levels of molar activity were determined (y axes). The relative activities against the various substrates are displayed above each bar, with the level of activity directed toward Leu-pNA (B and C, left) or toward Phe-␤NA (B and C, right) arbitrarily set to 100. ND, not detected. Data are presented as means with standard errors of results from two independent experiments. 43). Hence, we incubated recombinant LapA, with its His6 tag N terminal to its PA domain, along with supernatants (diluted 1/10) obtained from the wild-type 130b strain, the lspF mutant, or the proA mutant, and assayed for the cleavage of LapA by PAGE and Coomassie staining (Fig. 5A). Wild-type supernatants rapidly cleaved thẽ 44-kDa LapA, resulting in the appearance of an~35-kDa C-terminal portion, añ 13-kDa PA domain, and smaller cleavage/degradation products. However, proA mutant supernatants exhibited no evidence of cleavage over 2 h of incubation, appearing to be equivalent in that respect to the BYE medium control (Fig. 5A). The lspF mutant supernatants were also deficient for activity, although cleavage products did appear, most likely due to some release of ProA as a result of cell lysis. Compatibly with these data, purified LapA treated with (1/10-diluted) wild-type supernatants, but not proA mutant supernatants or BYE medium, cleaved methionine aminopeptide (Fig. 5B). Thus, we infer that ProA cleaves and activates LapA. Interestingly, when purified LapA was mixed for 1 h with undiluted rather than diluted supernatants, ProA-independent cleavage of LapA was seen (Fig. 5C). When immunoblots were used to examine the effect of undiluted proA mutant supernatants, the low-level cleavage event (the appearance of His-tagged PA) was detected after 5 min (Fig. 5D). However, undiluted proA mutant supernatants retained activity against leucine and methionine aminopeptides (Fig. 5E). Incidentally, the proA mutant supernatants still lacked activity against lysine aminopeptide, a LapB substrate (Fig. 5E), and ProA-dependent cleavage of LapB was observed on stained gels (Fig. 5F). Overall, these data showed that ProA is mainly responsible for activation of LapA but that LapA can also be cleaved, albeit to a lesser extent, by another secreted factor.
The structure of LapA. To determine how LapA provides its broad specificity, we began structural studies of recombinant LapA. Crystals were readily obtained, and its structure was determined by molecular replacement, using the coordinates of LapB (PDB code: 5GNE) (41) as the search model, and electron density maps were refined to 1.9 Å. The structure of LapA contains two molecules in the asymmetrical unit, and all residues could be built except for the N-terminal His tags and residues within the interdomain linkers (residues 82 to 101) that connect the PA domain (residues 1 to 81) and aminopeptidase domains (residues 102 to 378). The PA domain is formed of four helices and three ␤-strands, and the aminopeptidase domain has nine helices and seven ␤-strands. Although LapA was purified as a monomer, in our structure, the PA domain of one chain was associated with the aminopeptidase domain of an adjacent chain, which formed a domain-swapped dimer (Fig. 6A). The domain-swapped monomers were essentially identical, with a root mean square deviation (RMSD) over C ␣ atoms of 0.083 Å (without linkers), although variations were seen within the interdomain sequence. As an M28 family metalloprotease, the peptidase domain of LapA binds two Zn ions in its active site, which are coordinated by the conserved residues His194, Asp207, Glu241, Glu242, Asp269, and His347 and one water molecule (Fig. 6B). Drawing on structures of aminopeptidases from Streptomyces griseus (PDB code: 1F20) (44) and Aeromonas proteolytica (PDB code: 1XRY) (45), the main determinant of specificity in this class of enzyme is a binding pocket directly adjacent to the active site, and in LapA, this is formed from the side chains of Asp269, Met270, Phe285, Cys314, Cys318, Phe333, Cys335, Phe339, and His342 and from the backbone of Ser319, Thr337, and Ser338 (Fig. S4B). This creates a relatively deep hydrophobic cavity, which can readily accommodate aromatic and aliphatic residues ( Fig. 6C; see also Fig. S4B). In our structure of LapA, the PA domain covers the active site and substrate-binding pocket (Fig. S5) and thus represents the proenzyme form of LapA. A ring of predominantly hydrophobic residues on the surface of the enzymatic domain (Trp170, Leu198, Asp199, Tyr316, Phe339, His342, His347) provides a platform for the PA domain to dock, and this is mediated through its ␤-1 strand (Glu7), ␣2-helix (Ser39, Glu43, Thr46, Asp50), ␣2-␣3 linker (His56, Phe57, Asn59), and ␣3-helix (His62).
LapA promotes nutrient acquisition during infection. Given the enzymatic activity of LapA, which was revealed by structural analysis, we posited that LapA pro- for the indicated times with either BYE medium or 1/10-diluted supernatants (sup) obtained from late-exponential-phase broth cultures of wild-type strain 130b (WT), lspF mutant NU275 (lspF), or proA mutant AA200 (proA) and then processed by SDS-PAGE followed by Coomassie staining for total protein detection (A) or incubated with methionine aminopeptidase substrate in order to assess LapA activation (B). Arrows to the left of the gel image in panel A denote the migration position of FL-LapA, LapA missing its N terminus (ΔN-LapA), the PA domain of LapA (PA domain), and smaller cleavage products (PA domain=). The identity of the minor bands that were present within the input material (including one that appears to nearly comigrate with ΔN-LapA) and were maintained during incubation with the BYE medium or culture supernatants is unknown, but they likely represented breakdown products of the recombinant LapA. The migration of molecular mass markers (in kilodaltons) is also denoted to the left of the gel in panel A. Enzyme activity data in panel B are presented as means with standard errors of results from three independent experiments (n ϭ 3 each). Asterisks indicate significant increases in activity relative to t ϭ 5 min (*, P Ͻ 0.05 [Student's t test]). ND, not detected. (C and D) Purified FL-LapA was incubated at 37°C for the indicated times with either BYE medium or undiluted culture supernatants obtained from WT or proA mutant cultures and then processed for SDS-PAGE and Coomassie staining (C) or immunoblot analysis (IB) using anti-His antibodies directed against the hexahistidine tag embedded at the start of the PA domain (D). Arrows to the right side of the images in panels C and D denote the migration position of FL-LapA, ΔN-LapA, and the PA domain fragments, with the migration of the molecular mass markers denoted as well. Although the samples examined in panel D were on the same gel, they were not in adjacent lanes; therefore, we cropped out the intervening lanes that were not pertinent to the analysis. Data in panels C and D are representative of results of three and two experiments, respectively. (E) Filter-sterilized supernatants from late-exponential-phase broth cultures of wild-type 130b (WT), lapA mutant NU320 (lapA), and proA mutant AA200 (proA) were assayed for activity directed toward aminopeptidase substrates, with the N-terminal amino acid (AA) residue given. Changes in background-adjusted absorbance at OD 405 for the AA-p-nitroanilide substrates were monitored kinetically over 2 h. Data are presented as means with standard errors of results from three independent experiments (n ϭ 3 each). Asterisks indicate significant reduction in activity of a mutant relative to the WT (*, P Ͻ 0.05; **, P Ͻ 0.01; ***, P Ͻ 0.001 [Student's t test]). (F) Purified, full-length LapB (FL-LapB) (Input) was incubated at 37°C for the indicated times with either BYE medium or undiluted supernatants obtained from late-exponential-phase broth cultures of wild-type strain 130b (WT), lspF mutant NU275 (lspF), or proA mutant AA200 (proA) and then processed by SDS-PAGE followed by Coomassie staining for total protein detection. Arrows to the left of the gel image denote the migration position of FL-LapB and of LapB missing its N terminus (ΔN-LapB). The migration of molecular mass markers (in kilodaltons) is also denoted to the left of the gel. Data are representative of results from 3 independent experiments. motes L. pneumophila growth in A. castellanii by cleaving aminopeptides to release amino acids for uptake by the parasite. Thus, we asked if addition of a cocktail of amino acids to the Legionella-Acanthamoeba coculture could rescue growth of the lapA plaC mutant. To facilitate this experiment, we employed an amino acid cocktail (aspartate, alanine, glutamate, asparagine, glycine, serine, and proline) that was previously used to rescue the growth of a L. pneumophila mutant that was starved due to its lack of a T4SS effector (46). Thus, we used a reagent that was not inhibitory to L. pneumophila or its host. Moreover, the aspartate, alanine, and serine in the cocktail are predicted products of LapA activity (Fig. 4). Whereas amino acid supplementation had no effect on wild-type growth, it increased lapA plaC mutant numbers at 72 h postinoculation, and the results trended toward increases at 48 h (Fig. 7A). Thus, amino acid supplementa-  A. castellanii amoebas were infected with wild-type strain 130b (WT), lapA plaC mutant NU435, or lspF mutant NU275, and, at the indicated times, bacterial growth was determined. Where indicated, the medium in the infected wells was supplemented with 1 mM excess amino acids (AA) for the duration of the experiment. Growth data are presented as means with standard errors of results from three independent experiments (n ϭ 3 each). Asterisks indicate significant differences in CFU recovery between strains under the given conditions (*, P Ͻ 0.05; **, P Ͻ 0.01 [Student's t test]). ns, not significant. tion phenocopied the complementation results (Fig. 3C). The supplementation did not rescue the lspF mutant (Fig. 7B), which is compatible with this mutant lacking many exoproteins. Since neither the wild type nor the T2SS mutant was affected by addition of amino acids, the rescue of the lapA mutant was not a nonspecific effect on A. castellanii or L. pneumophila. Thus, the LapA mutant's infection defect is tied to its loss of enzyme activity. Given that LapA, PlaC, and amino acid acquisition does not fully account for the Ն1,000-fold defect observed upon complete loss of T2S (23), we sought additional secreted proteins that promote infection of amoebas.
Proteomic and mutational analyses identified NttD as a novel T2SS-dependent exoprotein that also promotes infection of A. castellanii. We next studied a 41-kDa protein that was previously detected in supernatants of L. pneumophila wild-type strain 130b grown in BYE broth but was absent in similarly derived supernatants of an lspF T2S mutant (28). Inspection of the 130b genome (47) revealed that the T2SS-dependent protein is encoded by open reading frame (ORF) L. pneumophila lpw10421, the second gene in a four-gene operon (Fig. S2C). Given our proteomics data and the bioinformatic analysis that is to follow, we named lpw10421 nttD, for novel type two secreted protein D, following the nomenclature used to denote nttA, nttB, and nttC (23,24). NttD is not annotated but is noted to have domain-of-unknown-function 4785. An unpublished crystal structure for NttD from L. pneumophila strain Philadelphia-1 has been deposited by others in the Protein Data Bank (PDB code: 4KH9). NttD is a mainly ␤-strand protein, and analysis using the PISA server (48) indicated that it exists as a dimer (Fig. S6A), although we detected equilibrium between the monomer and dimer in solution (Fig. S6B). Each chain is made of three unique domains, with the N-terminal domain mediating dimerization. While tertiary-structure analysis using the Dali server (49) detected no homologies for the N-terminal domain (no Z score of Ͼ8), the C-terminal domain shared features with glycosidases. Yet when we tested purified protein and concentrated supernatants against glycosidase substrates, no activity was seen. In order to discern a role for NttD, we made a mutant of 130b inactivated for nttD (NU431). The mutant grew in BYE broth like the parental 130b strain (Fig. S1A, right), indicating that NttD is not required for extracellular survival, at least under standard conditions. From examination of BYE supernatants, the mutant had normal levels of phosphatase, protease, and lipase activities (not shown), indicating that loss of NttD does not cause a general defect in T2S (24,50). Whereas the intracellular replication of the nttD mutant was not affected in the human U937 cell macrophages and Naegleria and Vermamoeba amoebas ( Fig. 8A to C), it was impaired for infection of A. castellanii (Fig. 8D). At 48 and 72 h postinoculation, the NU431-infected acanthamoebae yielded~30-fold-fewer legionellae. When infections were done in a transwell apparatus, such that bacteria were prevented from contacting host cells, the wild type and the mutant, though not growing, survived similarly (Fig. S1B, right), indicating that the reduced recovery of the mutant from the coculture was a result of defective infection rather than of impaired survival in the medium. When we tested a second nttD mutant (NU432) for infection of A. castellanii, the same defect was seen (not shown), implying that the loss of infectivity was due to the mutation in nttD rather than to a second-site mutation. Because a wild-type level of growth was achieved upon reintroduction of nttD into mutant NU431 (Fig. 8E), we concluded that NttD is required for infection of A. castellanii. Together, these data document NttD as a novel, T2SS-dependent exoprotein that promotes L. pneumophila infection of A. castellanii, thereby enhancing the range of Legionella in aquatic habitats. Discovering substantial roles for NttD, PlaC, and LapA in L. pneumophila infection of A. castellanii, we pondered to what degree these proteins account for the role of T2SS in infection. Thus, we tested a new mutant (NU441) that lacks lapA, plaC, and nttD. The triple mutant had a sizeable defect that went from~90-fold at 48 h postinoculation to~130-fold at 72 h (Fig. 8F). This defect was greater than that of the nttD mutant and lapA plaC mutant (Fig. 8F), indicating that NttD and LapA/PlaC operate in distinct pathways. The lapA plaC nttD mutant's defect did not fully recapitulate the lspF mutant's defect (Fig. 8F), indicating that there are additional T2S-dependent proteins that contribute to intracellular infection. However, the NttD, LapA, and PlaC data do explain, to a significant degree, the role of T2S in infection of A. castellanii.
Bioinformatic analysis revealed distinctive distribution patterns for nttD-, plaC-, and lapA-related genes within and outside the Legionella genus. All sequenced L. pneumophila strains and all of the 40 other sequenced species of Legionella have complete T2SSs (see Table S1 in the supplemental material). Therefore, we performed BLAST analyses to discern the species-and genus-wide distribution of nttD, plaC, and lapA. The nttD gene was in all L. pneumophila strains, and an nttD-like gene was in 36/40 Legionella species ( Fig. 9; see also Table S2A). On the basis of the distribution of nttD-like genes across the phylogenetic tree, nttD was found to be an ancestral gene that, though largely conserved, has been lost twice (Fig. 9). Further BLAST analysis found that NttD shares relatedness to hypothetical proteins in Ն9 genera of gammaproteobacteria and in one genus of deltaproteobacteria (Table 1). Similarly to nttD, plaC was conserved in the L. pneumophila species (Table S2B). A plaC-like gene was evident in 32 other Legionella species ( Fig. 9; see also Table S2B). The 8 remaining species lacked a plaC-like gene, indicating that the gene occurs in~80% of species. We surmise that plaC, like nttD, is an ancestral gene that, though reasonably conserved, has been lost twice (Fig. 9). Like nttD, plaC was lost prior to emergence of the adelaidensis-londiniensis-oakridgensis clade. Yet plaC appears to have also been lost prior to the advent of the feeleii-micdadei-drozanskii clade, resulting in a lower level of conservation than was determined for NttD (Fig. 9). Looking beyond Legionella, we found homologues of PlaC in genera of gammaproteobacteria, although these genera were different from those that had NttD homologues (Table 2). Interestingly, most non-Legionella homologues of PlaC, including the closest homologues, were in Cyanobacteria, a distant phylum that does not have a canonical T2SS (Table 2) (13). BLAST revealed that lapA occurs in all L. pneumophila strains (Table S2C). lapA-like genes were  Table S2C). Only L. maceachernii and L. londiniensis did not carry a lapA-like gene, due to two "recent" gene loss events. Thus, lapA is in~95% of species, indicating that it is an ancestral gene that is even more prevalent than nttD or plaC (Fig. 9). LapB, though conserved within the L. pneumophila species (Table S2D), is present in only 12% of non-pneumophila species ( Fig. 9; see also Table S2D). Seemingly, lapB is a recent gene acquisition, coincident with emergence of the clade containing L. pneumophila (Fig. 9). We also wondered about the prevalence of ProA, the T2SS-dependent protease that activates LapA and PlaC (Fig. 5) (35), and found that proA was 100% conserved (Fig. 9). Returning to the distribution of LapA-like FIG 9 Distribution of genes encoding key type II-secreted proteins within the Legionella genus. Gain and loss events pertaining to genes encoding type II-secreted NttD, LapA, LapB, PlaC, and ProA were modeled on a maximum-likelihood tree of 41 Legionella species, which was constructed based on the protein sequence encoded by 78 nearly universal genes (9). Colored circles and arrows indicate a gene gain event (upward-pointing arrow) or loss event (downward-pointing arrow) based on maximum parsimony, and colored cells (right) indicate the presence of nttD, lapA, lapB, plaC, and proA within a given species. Analysis of non-Legionella homologues was done with the exclusion of hits from the Legionella genus only with homologues predicted using an E-value cutoff of Ͻ4 ϫ 10 Ϫ10 and an alignment length of at least 65% (80). b Only the top representative species per genus is given, based on the closest homology with NttD. c I, percent identity; S, percent similarity; C, percent coverage. proteins but looking beyond Legionella, BLAST analysis found that LapA's closest prokaryotic homologues occur in betaproteobacteria (Table 3). This pattern of distribution of non-Legionella homologues was very different from that of NttD and PlaC, which included primarily gammaproteobacteria or cyanobacteria and no betaproteobacteria (Tables 1 and 2). Even more interestingly, the closest relative, which had 41% identity and 60% similarity with LapA, was an aminopeptidase of A. castellanii (Table 3). Other eukaryotic LapA-like proteins were in fungi, and in many cases, they were closer to LapA than the prokaryotic homologues (Table 3). Thus, LapA has strong eukaryoticprotein-like traits.

DISCUSSION
The experimental data presented increase our understanding of T2S and hostparasite interactions in multiple ways. With the definition of NttD and LapA as major potentiators of infection of A. castellanii, the number of T2S effectors confirmed as being involved in infection of amoebas rises to seven (23,24,30,31). Thus, approximately one-third of the known effectors are required for infection (13), implying that  intracellular parasitism has played a large role in shaping the T2SS repertoire. That NttD, along with the previously described NttA and NttC, has a role in amoebal infection and yet lacks similarity to known enzymes suggests that many T2SS substrates may be specialized for the intra-amoebal lifestyle (23,24). Although aminopeptidases in the M28 family have been linked to T2S and/or virulence (37,51,52), the results from this study of LapA mark the first documentation of a secreted aminopeptidase promoting an intracellular infection event. Thus, prokaryotic LapA-like proteins (Table 3) may be facilitating survival for bacteria such as Duganella and Burkholderia spp. that reside in amoebas (53). In establishing a link between LapA and PlaC, we have also revealed a unique example of functional redundancy operating during infection. While promoting growth in A. castellanii, NttD and the combination of LapA and PlaC were not needed for infection of other amoebas. This highlights how the importance of T2SS substrates varies in a host-specific fashion; presumably, each amoebal type represents a distinct environment and, hence, L. pneumophila invokes different exoproteins in order to efficiently establish its host range. Unlike NttD and LapA, PlaC was critical for infection of every amoebal host. There are probably commonalties across different amoebas, and so L. pneumophila repeatedly relies upon some of its effectors to flourish in aquatic habitats. The finding that ProA processes LapA and LapB adds to past work showing that ProA cleaves PlaA and PlaC and is required for strain 130b infection of V. vermiformis and N. lovaniensis (30,35,43). Another study found that a proA (mspA) mutant of strain JR-32 is impaired for infection of A. castellanii (19). Thus, ProA rivals PlaC in its broad significance. Although previous studies reported on the transcription of a few T2S genes (54)(55)(56), the current analysis represents the largest assessment of T2S effector expression by L. pneumophila. After examining 19 genes in two stages of extracellular growth and in three hosts, it is apparent that secreted-protein genes exhibit a variety of expression patterns rather than showing similar responses to conditions. From this, we intuit that the amounts of secreted protein made are controlled at the level of the individual or of subsets of effector-gene transcription more than at the level of apparatus gene transcription. Bioinformatic analyses revealed further insights into T2SS and the origins of its effectors. Being found in the majority of Legionella species, the genes encoding NttD, PlaC, and LapA appear to be ancestral genes that underwent only a few loss events during the evolution of the genus. On the basis of the dispersal of homologues in other types of organisms, ancestral NttD was likely acquired from related gammaproteobacterial species, PlaC from cyanobacterial species, and LapA from protozoa or fungi. ProA is encoded by another ancestral T2SS gene, but in this case, no loss events were detected. Unlike NttD, PlaC, LapA, and ProA, LapB was only in the "recent" clade containing L. pneumophila. We posit that LapB arose from lapA duplication. This is supported by phylogenetic analysis of LapA and LapB, which form a monophyletic group exclusive of other bacteria or eukaryotes, and is based on the fact that LapB has acquired substitutions relative to the ancestral LapA (Fig. 10). Given that NttD, PlaC, LapA, and ProA promote infection but LapB does not, we speculate that ancestral, highly conserved T2SS effectors have the greater role in intracellular parasitism. This evolutionary pattern for T2SS effectors differs from that of most T4SS effectors, which are distributed in Ͻ25% of Legionella species (9).
The activities associated with PlaC and LapA provide hypotheses for how these T2S substrates promote infection. PlaC not only generates fatty acids but also can modify lipids within amoebal vacuoles (34,35). Thus, PlaC may promote intracellular nutrient acquisition via direct lipid metabolism, or through modification of lipids in LCV membranes, which might in turn influence vesicular trafficking or anchoring of other effectors. LapA likely promotes replication through its ability to generate amino acids that are limiting during the intracellular lifecycle. Our analysis found a broad spectrum of activity for LapA, and we showed that amino acid supplementation rescues the growth of mutants lacking LapA. Interestingly, the leucine, methionine, isoleucine, and valine generated by LapA, coupled with arginine produced by LapB (30), account for all of the amino acids that cannot be synthesized by L. pneumophila or A. castellanii (46).
It is intriguing to contemplate why LapA and PlaC, two very different exoenzymes, should overlap in terms of function. In one scenario, L. pneumophila utilizes both amino acids (generated by LapA) and lipids (generated by PlaC) for nutrition within the LCV. Thus, a growth defect is revealed only when both LapA and PlaC are absent, depriving the bacterium of two food sources. Alternatively, changes to LCV membranes mediated by PlaC might promote the import of nutrients beyond those targeted by LapA. Hence, without PlaC and LapA, intravacuolar bacteria again lose two food sources. To discern how PlaC and LapA operate, it would be helpful to learn their locations in the infected cell. Proteins secreted by T2S, including ProA, can be translocated out of the LCV (57), and thus, it is possible that LapA or PlaC or both have targets in the host cytoplasm. Overall, LapA and PlaC modulate different processes and yet accomplish a single task.
With the characterization of LapA, we have a striking example of a eukaryoticprotein-like, T2SS-dependent exoprotein, indicating that eukaryotic-protein-like effectors of L. pneumophila are not limited to the T4SS (58). LapA was likely acquired through horizontal gene transfer within A. castellanii or an ancestral protozoan. Indeed, the top-scoring LapA homologue outside Legionella belongs to A. castellanii and is a putative aminopeptidase. Also, LapA shares higher-than-average homology to its topscoring eukaryotic homologue (41% identity, 60% similarity), compared to the average levels of identity and similarity of all other (primarily T4SS) eukaryotic-protein-like  Table 3 are noted in parentheses after the respective genus designations. proteins of L. pneumophila (34% identity, 52% similarity) (59). Moreover, phylogenetic reconstruction of aligned amino acid sequences of LapA homologues revealed that LapA is not descended from other bacteria but shares a common ancestor with A. castellanii (Fig. 10). Interestingly, the A. castellanii aminopeptidase is secreted (60). Thus, legionellae benefit from secreting a host-like aminopeptidase as a means of scavenging limiting amino acids.
The structure of LapA is similar to those of several known aminopeptidases. The Dali server identified LapB (41) and A. proteolytica Apap (PDB code: 1AMP) (61) as having the highest tertiary homology (Z scores of 54.6 Å and 45.0 Å, respectively), although this is concentrated at the peptidase domain (see Fig. S7A in the supplemental material). For example, the positions of active-site residues in LapA and LapB structures are almost identical (Fig. S7B), while there is a significant reorganization of helices between PA domains. This accommodates differences in their substrate-binding pockets, with LapB's cavity being wider and more negatively charged (Fig. S7C), compatible with a preference for positively charged lysine and arginine (30). Often, the role of the PA domain is to keep an enzyme in an inactive state prior to secretion (62), although for purified LapA, it did not prevent detection of activity. The effect of the PA domain is likely due to a transient association between the PA and peptidase domains in LapA, as we observed a monomeric and domain-swapped form in solution. Although domain swapping was not seen in the structure of LapB, it dimerizes in solution and its PA domain does not prevent detection of activity under some conditions (41). The LapA structure is the fifth structure for a Legionella T2SS effector, joining those of NttD, LapB, phosphatase Map, and VirK-like Lpg1832 (28,41,(63)(64)(65).
Although the PlaC, LapA, and NttD data explain, to a large extent, the impact of T2S on L. pneumophila infection of A. castellanii, future work should further define the Legionella T2SS output. Proteomic assessment of supernatants and subsequent (single) mutant analysis have heretofore been the means to identifying T2SS-dependent proteins that promote infection (13). These approaches helped reveal the nature of NttD as a potentiator of infection and will continue to offer new targets for analysis (28). Yet this strategy clearly failed to reveal the roles that LapA and PlaC have in A. castellanii infection (23,30), undoubtedly due to the redundancy that exists between LapA and PlaC. The new-found importance of these two effectors was uncovered through a combination of transcriptional and mutational analyses. This approach has four steps: (i) identification of the genes that are upregulated during wild-type infection of the host; (ii) examination of the transcript profile of mutants that lack the upregulated genes, while being alert to compensatory increases in other transcripts; (iii) mutation of two or more upregulated genes that exhibit a compensatory link; and (iv) testing of mutants for loss of infection. Our data set predicts more substrates that may promote infection; e.g., two strongly upregulated genes during infection of A. castellanii and N. lovaniensis were legP encoding a putative protease and map encoding a phosphatase ( Fig. 1B and C). Though neither a legP mutant nor a map mutant was impaired for infection (23), a legP map mutant might be defective. Interestingly, LegP and Map, like LapA, have eukaryotic-protein-like homologs (23,28,64).

MATERIALS AND METHODS
Strains, plasmids, and media. L. pneumophila 130b (ATCC BAA-74) served as the wild type and parent for all mutants (25). proA mutant AA200, lspF mutant NU275, lapA mutant NU320, lapB mutant NU322, lapA lapB mutant NU324, celA mutant NU353, and plaC mutants NU367 and NU420 were described before (14,15,23,29,30,66). Newly derived mutants lacking nttD, lapA, or plaC were made using methods of allelic exchange (23,25). Complemented mutants were obtained by introducing intact copies of nttD, lapA, or plaC on a plasmid, analogously to past studies (24). Primers and plasmids used in mutant and complement constructions are listed in Table S3 in the supplemental material. Legionellae were routinely grown at 37°C on buffered charcoal yeast extract (BCYE) agar or in BYE broth (25  Ramachandran analysis e Residues in outlier regions (%) 0.0 Residues in favored regions (%) 98.6 Residues in allowed regions (%) 100 a Numbers in parentheses refer to the outermost resolution shell. b R sym ϭ Α|I Ϫ ϽIϾ|/ΑI, where I is the integrated intensity of a given reflection and ϽIϾ is the mean intensity of multiple corresponding symmetry-related reflections. c R work ϭ Α||F o Ϫ |F c ||/ΑF o , where F o and F c are the observed and calculated structure factors, respectively. R free ϭ R work values calculated using 10% random data excluded from the refinement. d The RMSD stereochemistry value represents the deviation from ideal values. e Ramachandran analysis was carried out using Molprobity (81). the release of p-nitroanilide (p-NA) from L-leucine p-NA, L-lysine p-NA, L-methionine p-NA, glycine p-NA, L-aspartate p-NA (Chem-Impex), L-isoleucine p-NA (Chem-Impex), L-proline p-NA (MP Biomedicals), L-glutamate p-NA (Chem-Impex), and L-valine p-NA (Chem-Impex), respectively (30). Phenylalanine, alanine, and serine aminopeptidase activities were measured by release of ␤-naphthylamide (␤-NA) from L-phenylalanine ␤-NA, L-alanine ␤-NA, and L-serine ␤-NA (Chem-Impex), respectively (30). Purified LapA-FL and LapA-ΔN were diluted to a concentration of 500 nM in the reaction buffers and assayed as described above. One unit of activity is equal to that which catalyzes the conversion of 1 mol substrate/min, and molecular activity was expressed as the number of units per micromole of purified enzyme, based on standard curves corresponding to p-NA and ␤-NA (30,69). Assays for ␣-mannosidase, ␤-mannosidase, ␤-galactosidase, and endo-1,4-␤-mannanase were as previously described (70)(71)(72).
Assays for the cleavage, activation, and inhibition of aminopeptidases. To assay for the ability of secreted proteins to cleave aminopeptidases, 10 g of purified LapB-FL or LapA-FL was incubated at 37°C for up to 2 h with 50 l of 1ϫ or 1:10-diluted late-log-phase culture supernatants, respectively. Samples (10 l) were collected on ice and were then subjected to SDS-PAGE. Protein was visualized with SimplyBlue SafeStain (Thermo Scientific). Immunoblots were done as previously described (17), with recombinant protein detected with mouse anti-His (Millipore; clone HIS.H8). For real-time monitoring of LapA-FL cleavage and resultant activation, LapA-FL (500 nM) was added to the assay buffer containing 3 mM p-NA-based substrate and the supernatant (1:10 final concentration). Absorbance was monitored at 405 nm. LapA inhibition assays were done using 500 nM LapA-ΔN incubated with serial dilutions of bestatin. Residual activity following treatment was determined based on release of p-NA from L-leucinep-NA as described above.
Crystal structure determination. Crystallization of LapA (10 mg/ml) was done using the sitting-drop vapor-diffusion method and growth at 293 K in 0.15 M ammonium sulfate-0.1 M 4-morpholineethanesulfonate (pH 5.5)-25% (wt/vol) polyethylene glycol 4000. Crystals were soaked for 30 to 60 s in cryoprotection solution (well solution supplemented with 20% glycerol) and flash-cooled in liquid nitrogen. Diffraction data were collected at 100 K on Beamline I03 at the Diamond Light Source (DLS), United Kingdom. Data were processed using XDS (73) and scaled using AIMLESS (74). Initial phases were obtained using PHASER (75) with the structure of LapB (41) as the search model. Density modification was done with PARROT (76) followed by automated model building with ARPWARP (77). The remaining structure was manually built within COOT (78), and refinement was carried out with refmac (79) using noncrystallographic symmetry and translation-libration-screw groups. Ten percent of the reflections were omitted for cross-validation. Processing and refinement statistics for the final model are given in Table 4.
Accession number(s). Coordinates and structure factors have been deposited in the wwPDB under PDB code 6ESL.