NAD(H)-mediated tetramerization controls the activity of Legionella pneumophila phospholipase PlaB

Since PlaB is a virulence factor that raises potential biosafety concerns when produced in a heterologous host, we employed the inactive D203N mutant in our extensive efforts to obtain high-quality crystals. Most crystals suffered from anisotropic diffraction and streaky, overlapping reflections. The structure was finally determined at 2.3-Å resolution by single-wavelength anomalous dispersion of seleno-L-methionine–labeled PlaB in a crystal obtained by seeding (SI Appendix, Fig. S1 A and B). This crystal belonged to space group P2₁ and displayed strong translational noncrystallographic symmetry, explaining why the refinement converged at very high R-factors for this data set (SI Appendix, Table S1). The asymmetric unit comprised four chains that arrange in a tetrameric fashion, best described as a dimer of dimers (Fig. 1A). Interestingly, the interface between both dimers is relatively small, and PISA analysis of the protein chains alone did not recognize this tetramer as a stable assembly (20). The PlaB dimers, however, are predicted as stabilized by approximately −29.8 kcal/mol, suggesting that they represent the active form of PlaB. The dimers are formed by head-to-tail interaction between the N-terminal phospholipase domain of the first monomer and a protruding hook-like extension at the C terminus of the second monomer (residues R446 to D474) (Fig. 1 A–C and SI Appendix, Table S2). This leads 1) to structural complementation of the central β-sheet of the phospholipase domain by β-18 from the hook region (see Fig. 3B), explaining the previously determined loss of activity when residues from the C terminus are deleted (15, 16), and 2) to a number of polar and hydrophobic interchain interactions between the last ∼30 amino acids and residues of the N terminus.

The PlaB monomer consists of two clearly distinguishable domains (Fig. 1 C and D and SI Appendix, Fig. S2). These comprise residues 1 to ∼324 and residues ∼325 to 474, representing the N-terminal phospholipase domain and a CTD consisting of a bilobed β-sandwich extended by the hook structure, respectively. The NTD displays a typical α/β-hydrolase (ABH) fold, hallmarked by a central six-stranded parallel β-sheet (spatially ordered as β-2, β-1, β-3, β-4, β-5, β-8) surrounded by 17 α-helices (α-1 to α-17) (Fig. 1 A, C, and D) (21, 22). The catalytic center is readily discernible by the catalytic triad S85/D203/H251 that has previously been identified through sequence alignments and mutagenesis (Fig. 1 B and C) (15). It is shielded from the solvent by a closed lid formed by residues G127 to G144 of α-8 and α-9, as is frequently observed in ABHs in the absence of substrates (Fig. 1 B and C and SI Appendix, Table S2) (23). The quality of the electron density indicates that this lid is highly flexible. In addition to these canonical features, the NTD also possesses unique structural elements that are specific to PlaB, namely a long α-helix (α-17, residues R285 to N305) and two antiparallel β-sheets (β-6/β-7, residues S216 to R237 and β-9/β-10, residues K308 to I321) (Fig. 1 A, C, and D and SI Appendix, Table S2). While the α-helix participates in lining the large central β-sheet, the additional β-sheets β-6/β-7 and β-9/β-10 protrude from the ABH fold (Fig. 1 A and C and SI Appendix, Fig. S2).

The CTD of PlaB is dominated by a bilobed β-sandwich consisting of a six-stranded mixed (β-6, β-7, β-11, β-13, β-14, and β-17) and a three-stranded antiparallel β-sheet (β-12, β-15, and β-16) (Fig. 1 C and D and SI Appendix, Fig. S3). Interestingly, two strands of the first β-sheet stem from the NTD (β-6 and β-7) (Fig. 1 C and D). Searches with DALI (24) reveal similarity to immunoglobulin-like domains; however, none of the identified structures possesses the mosaic nature and the ensuing β-sheet extension by β-6/β-7, suggesting that the CTD of PlaB represents a not previously observed fold (SI Appendix, Fig. S3).

PlaB hydrolyzed a broad array of biologically important phospholipids and lysophospholipids. Activity was found toward negatively charged substrates such as PG, phosphatidylserine, phosphatidylinositol (PI), and also toward PC, which contains a positively charged head group (Fig. 2A). With respect to PI, different derivatives were hydrolyzed (SI Appendix, Fig. S4). PIs are pivotal constituents of the LCV that impact on its membrane dynamics and on vesicle trafficking. L. pneumophila modifies these compounds to its own advantage during host infection (25). To gain insight into determinants for activity and this substrate spectrum, we analyzed the importance of certain structural elements, specifically the lid, the hook, β-6/β-7, and β-9/β-10. A lid is a typical feature of lipases regulating substrate access to the catalytic site. Both its amphipathic nature and specific amino acids are crucial for activity and specificity (23, 26). Indeed, we found that the lid point mutation S129A had severely reduced activity toward PC but only slightly reduced activity toward PG and lysophosphatidylglycerol (LPG) (Fig. 2 B–D). This is in accordance with previous findings showing that this residue, now recognized as part of the lid, promotes PC hydrolysis and hemolysis (15). Furthermore, the positive charge of the neighboring lid residues R130 and R133 could contribute to the recognition of negatively charged lipid substrates, as has been observed for other lid-containing hydrolases (21, 23). Indeed, whereas point mutations led to reduced activity toward PG and LPG but also influenced PC hydrolysis, the triple mutant S129A/R130A/R133A was inactive toward all substrates tested (Fig. 2D). These findings demonstrate that the lid region influences activity and substrate specificity.

The noncanonical β-strands β-6/β-7 and β-9/β-10 are antiparallel to each other and project away from the ABH fold (Fig. 1 A and C and SI Appendix, Fig. S2). Given the importance of strands β-6 and β-7 within the structure of the CTD (SI Appendix, Fig. S3), it is not surprising that the respective deletion mutant was inactive (SI Appendix, Fig. S5 A and B). A similar phenotype was found when the hook (residues R446 to D474) was deleted (SI Appendix, Fig. S5 A and B). This is in accordance with previous observations where shortening of the CTD led to a decrease in activity (16), corroborating our finding that the hook connects two PlaB monomers and extends the ABH fold to yield active PlaB dimers. The Δβ-9/β-10 mutant lost its activity against PC but retained a lower level of activity toward PG and LPG. Therefore, β-9/β-10 affects the substrate spectrum of PlaB (Fig. 2 E and F). An intriguing pattern of hydrophobic and cationic residues displayed on both sides of the β-9/β-10 sheet (K308, F310, K312, F316, R318, and Y320) resembles motifs known to contribute to cation–π interactions in proteins that bind to positively charged phospholipid head groups (Fig. 2E) (27, 28). Consistently, the corresponding triple mutant F310D/F316D/Y320D indeed resembled the Δβ-9/β-10 mutant in terms of activity (Fig. 2 E and F).

The position of sheet β-9/β-10 in the tetramer suggests that it may be involved in dimer/dimer interactions (Fig. 1A) and indeed, as size-exclusion chromatography-multiangle light scattering (SEC-MALS) experiments showed, deletion or mutation impeded concentration-dependent tetramerization. Specifically, whereas different concentrations of PlaB wildtype (WT) revealed either the tetramer only or a tetramer-dimer mixture, the Δβ-9/β-10 or F310D/F316D/Y320D mutants were present in a dimeric form regardless of protein concentration (Fig. 3A and SI Appendix, Table S3). Together, these data establish that β-9/β-10 not only tunes selectivity of PlaB toward PC but also contributes to stabilization of the tetrameric state.

Furthermore, because of its exposed location, it is conceivable that, when the tetramer deoligomerizes to active dimers, sheet β-9/β-10 is liberated and may subsequently be involved in OM association. We therefore investigated association of PlaB and the Δβ-9/β-10 mutant to PG and PC liposomes since PG and PC are major constituents of the L. pneumophila OM (29). We found that higher amounts of Δβ-9/β-10 than of PlaB WT remained in the supernatant, suggesting that it associated less efficiently to the liposomes (Fig. 3B). Similar results were observed upon addition of the proteins to the L. pneumophila ΔplaB mutant. Specifically, higher amounts of Δβ-9/β-10 were present in the supernatant and accordingly lower amounts associated with the bacteria (Fig. 3C). This indicates that β-9/β-10 is an important element acting as membrane attachment structure, a fact which might additionally contribute to efficient phospholipid hydrolysis (Fig. 2F).

Next, we were interested if OM localization of PlaB was affected by deletion of β-9/β-10. We first confirmed membrane, OM, and surface localization of PlaB WT in L. pneumophila by means of 1) subcellular fractionation (SI Appendix, Fig. S7 A and B), 2) proteinase K digestion (SI Appendix, Fig. S7C), and 3) surface-protein biotin labeling (SI Appendix, Fig. S7D). Since the export mechanism of PlaB in L. neumophila is unclear and did not involve several Legionella-specific secretion systems (13), we tested if it is exported in Escherichia coli. In addition, implementation of E. coli as a model system would further allow the analysis of PlaB variants released from the membrane in the absence of potentially PlaB-degrading enzymes, such as the L. pneumophila zinc metalloproteinase ProA (30). We indeed found that PlaB and the catalytic mutant D203N localized to the OM (Fig. 3D and SI Appendix, Fig. S8). However, the Δβ-9/β-10 and the F310D/F316D/Y320D mutants showed subcellular mislocalization, specifically a main protein fraction associated with the inner membrane (Fig. 3D and SI Appendix, Fig. S9). All other mutants in the lid, hook, or β-6/β-7 localized to the OM similar to the WT protein (SI Appendix, Figs. S8 and S10). In addition, Δβ-9/β-10 was released into the culture supernatant, unlike intact PlaB (Fig. 3E). To further substantiate our findings, we expressed truncations of PlaB in L. pneumophila and analyzed surface presentation by means of proteinase K digests, which again corroborated the importance of β-9/β-10 (SI Appendix, Fig. S11). In summary, the Δβ-9/β-10 protein was malfunctional in terms of tetramer stabilization and in OM association. Since these two aspects are important for control and localization of PlaB activity, Δβ-9/β-10 may have detrimental effects for the bacteria. Indeed, a L. pneumophila ΔproA strain expressing Δβ-9/β-10 released isocitrate dehydrogenase (ICDH), a cytosolic marker, but not a strain with WT PlaB (Fig. 3F). Furthermore, a reduction of Δβ-9/β-10 in the cell pellet was observed at higher ODs, indicating membrane release of the variant. However, Δβ-9/β-10 was not detected in the culture supernatant likely because of proteolytic degradation by enzymes other than ProA. In conclusion, this confirms the key importance of the β-9/β-10 structural element in functional and spacial control of L. pneumophila PlaB.

The fact that the tetramer is not recognized as a stable entity in structure-based calculations with PISA yet is a dominant species in experiments raises the question how it may be stabilized in solution. While tetramer formation was found to be concentration dependent, we were surprised to observe copurified ligands at the dimer/dimer interface, albeit at low occupancy in the initial crystal structure. The shape of the electron density suggested the identity of these ligands as NAD(H) molecules that bind to two different but closely neighbored binding sites per monomer (Fig. 4 A and B). Tetramers were the only species found to bind NADH here. In detail, we observed that 1) the presence of NAD(H) in the SEC-MALS running buffer led to the exclusive formation of tetramers (Fig. 4C), 2) in freshly purified PlaB, the A260/A280 ratio in the SEC-MALS experiment was higher than expected for the peak corresponding to tetrameric PlaB but not for lower oligomers (SI Appendix, Fig. S12), 3) the addition of NAD(H) led to a more defined melting point (SI Appendix, Fig. S13), and 4) thio-NAD (SNAD), a component of the thermal shift assay (TSA) used to optimize the storage buffer for PlaB, improved and accelerated crystallization of the PlaB tetramer, yielding crystals in spacegroup P1 that diffracted up to 1.8 Å and that grew to full size after 20 instead of 150 d (SI Appendix, Fig. S1C). The corresponding structure indeed revealed full occupation of the eight NAD(H) binding sites with SNAD molecules (Fig. 4A and SI Appendix, Fig. S14). The binding sites seem to be accessible through large channels in the dimer/dimer interface (SI Appendix, Fig. S15), and several residues are involved in specific interactions with the ligand, particularly from the NTD (β-9/β-10, α-13, and lid) of one PlaB dimer and from the CTD (β-12, β-13, and β-14) of the other, also including π-stacking between the nicotinamide group and Y190 and Y196 (Fig. 4B). Interestingly, while all eight NAD(H) binding sites are occupied with SNAD, a substantial part of the lid of all four monomers became extremely flexible such that residues 133 to 141 (RIKSFFEGI) could not be traced in the corresponding structure. The noncanonical strands β-9/β-10 of one dimer line the outer NAD(H) binding site, and R318 is involved in a hydrogen bond with the ribose unit of the nicotinamide half of the ligand (Fig. 4B and SI Appendix, Fig. S14).

When we reanalyzed the crystal structure of PlaB in the ligand-bound form with PISA, we found that the tetramer was calculated to be strongly stabilized by SNAD (ΔG = −44.7 kcal/mol with SNAD and −3.0 kcal/mol without SNAD) (20). This suggests that tetramerization and hence inactivation of PlaB is NAD(H) mediated. Consistently, we confirmed that NAD⁺ and NADH inhibited the activity of WT PlaB (Fig. 4D). Although residues of β-9/β-10 among others were found in close proximity to the ligand, the activity of the Δβ-9/β-10 and the F310D/F316D/Y320D mutant was also reduced by NAD(H), demonstrating the importance of additional NAD(H) interaction sites (Fig. 4 A, B, and D). Furthermore, it was possible to reconstitute tetramers for all concentrations of the WT and mutant proteins when NAD(H) was added (Fig. 4C). This shows that the tetrameric form of the mutants, although less stable without ligand addition (Fig. 3A), was stabilized by the ligand (Fig. 4C). We conclude that NAD(H) enhances PlaB tetramerization and concomitantly enzyme inactivity, establishing an allosteric regulatory process.

We used the crystal structure to design mutants that should have impaired NAD(H) binding. We mutated R130, R366, and Y378 that tightly interact with NAD(H) in one or both binding sites and combined these with each other or with Δβ-9/β-10. However, combined mutants lost activity and single mutants which retained at least partial activity still responded to NAD(H) (SI Appendix, Fig. S16). This led us to the conclusion that deletion of NAD(H) binding sites will also impair activity.

A BLAST search revealed high conservation of PlaB in the Legionella genus, mostly showing over 80% protein identity with the exception of about one-third of the non-pneumophila species (SI Appendix, Fig. S17) (31). PlaB-like proteins were further found in a variety of genera of several bacterial phyla. Using the UniRef90 database (32), we constructed a sequence similarity network of about 200 PlaB-like proteins by defining clusters of 50% sequence identity (SI Appendix, Fig. S18) (33). Alignment of representatives of these clusters revealed that the two-domain structure of PlaB seems preserved. The β-9/β-10 element and NAD(H)-interacting residues were less conserved (SI Appendix, Fig. S19). We conclude that PlaB is widely distributed in bacteria and that residues for NAD(H) interaction and OM association via β-9/β-10 are highly conserved in many Legionella species, suggesting a similar control mechanism which may further be relevant in some other bacteria.

Bacterial strains, plasmids, and primers are listed in SI Appendix, Tables S4 and S5. L. pneumophila strain Corby was used for plaB cloning (49). Versions of PlaB were generated by means of the QuikChange Site-Directed Mutagenesis Kit (Stratagene) (SI Appendix, Table S4). Primers were obtained from Eurofins MWG Operon and IDT (SI Appendix, Table S5). The bacterial strains were cultivated in lysogeny broth media containing 100 µg/mL ampicillin or 30 μg/mL chloramphenicol (for E. coli) or buffered yeast extract (BYE) broth when required containing 6 μg/mL chloramphenicol (for L. pneumophila).

PlaB for crystallization was produced as N terminally Strep-tagged inactive Strep-PlaB D203N as described previously (16). Briefly, protein was expressed in E. coli BL21 (pKK21), using 50 µM isopropyl 1-thio-D-galactopyranoside (IPTG) for induction. Seleno-L-methionine labeling was achieved by growing the bacteria in M9 minimal medium supplemented with 50 mg/L seleno-L-methionine (SeMet) and inducing with 50 µM IPTG overnight at 20 °C. Purification involved Strep-Tactin affinity and size-exclusion chromatography, followed by concentration to 10 mg/mL and storage at −80 °C until further usage. For activity and localization analyses, PlaB and variants were expressed using 0.1 mM IPTG for induction and purification by means of Strep-Tactin affinity and size-exclusion chromatography.

SEC-MALS experiments were performed on a chromatography system equipped with a Superdex 200 Increase 10/300 GL SEC column, a miniDAWN TREOS MALS detector, and an Optilab T-rEX refractometer (Wyatt Technology Corp.) The column was equilibrated with 150 mM NaCl, 50 mM Tris (pH = 8), and different volumes of protein solution at 10 mg/mL were analyzed. Data were processed with the Astra software package (Wyatt Technology Corp.)

TSAs were performed by mixing 5 μL protein solution at 2 mg/mL with 5 μL SYPROOrange (1:50 diluted from 5,000× stock solution, Invitrogen), 5 μL eightfold buffer (1,200 mM NaCl and 400 mM Tris pH = 8), and 35 μL RUBIC Additive Screen (Molecular Dimensions) in a 96-well plate. Data were collected with a C1000 Touch Thermal Cycler, equipped with a CFX96 Real-Time System from 4 to 95 °C in 1 °C increments. TSA data were analyzed with the software BIO-RAD CFX96 Manager 3.1 (BIO-RAD).

Initial crystallization conditions were identified with the vapor diffusion method in sitting drops consisting of 200 nL protein solution at a concentration of 1 to 6 mg/mL mixed with the same volume of precipitant and equilibrated against 60 µL reservoir at 20 °C. Precipitants yielding crystals were optimized in grid and random screens. Microseeding experiments were performed with an OryxNano liquid dispenser (Douglas Instruments) using 100 nL seed stock of native PlaB crystals broken up in 500 µL of the corresponding reservoir and 300 nL SeMet PlaB at 4 mg/mL mixed with 200 nL precipitant solution. Crystals were further optimized by preincubating the protein with 1 mM thio-nicotinamidedinucleotide (SNAD). Diffraction data collection proceeded with crystals obtained with precipitants consisting of 100 mM Tris (pH = 8.5), 8.33% glycerol, 10.8% 2-propanol (native PlaB) and 77.8 mM NaSCN, 1.67% glycerol, 5.44% PEG400, 11.6% 2-propanol, and 4.11% tacsimate (SeMet PlaB). Crystals were cryoprotected in mother liquor supplemented with 10% (vol/vol) (2R,3R)-(−)-2,3-butandiole before flash cooling in liquid nitrogen. Diffraction data were collected on beamlines BL14.2 [synchrotron BESSYII, Helmholtz Zentrum Berlin (50)], P11 [synchrotron PETRAIII, Deutsches Elektronen-Synchroton (DESY) (51)] and PXIII/X06DA [synchrotron Swiss Light Source (SLS), Paul Scherrer Institute (52)]. Indexing and integration was achieved with XDS, and scaling involved AIMLESS or STARANISO (53–57), as summarized in SI Appendix, Table S1.

Initial phases were derived from single anomalous dispersion differences in diffraction data of a crystal of SeMet-labeled PlaB collected at the K-absorption edge of selenium. Heavy atom positions were identified with hkl2map (55) and forwarded to phenix.phaser of the PHENIX software suite (58). An initial model was obtained with phenix.autobuild and completed manually in Coot (59) from the CCP4 software suite (56). Further refinement involved alternating rounds of manual adjustments and optimization in phenix.refine. Model statistics are provided in SI Appendix, Table S1. Coordinates and structure factor amplitudes have been deposited in the Protein Data Bank (60) under accession code 6zth [seleno-L-methionine–containing structure with low-occupancy NAD(H)] and 6zti (unlabeled PlaB–containing SNAD at full occupancy). Molecular structures are displayed with PyMOL (61).

Purified protein, bacterial lysates, or culture supernatants of E. coli BL21 expressing PlaB or its versions were incubated with different lipids (dipalmitoyl phosphatidic acid [PA], dipalmitoyl phosphatidylcholine [PC], dipalmitoyl phosphatidylglycerol [PG], dipalmitoyl phosphatidylethanolamine [PE], dipalmitoyl phosphatidylserine [PS], 1-monopalmitoyl-lysophosphatidic acid [LPA], 1-monopalmitoyl-lysophosphatidylcholine [LPC], 1-monopalmitoyl-lysophosphatidylglycerol [LPG], 1-monopalmitoyl-lysophosphatidylethanolamine [LPE] [Avanti Polar Lipids, Inc.], and dipalmitoyl phosphatidylinositol [PI] [MoBiTec]), 3 mM NaN₃, 0.5% Triton X-100, and 20 mM Tris HCl (pH = 7.2) as described previously. Prior to incubation, the lipids were vortexed for 15 min at 37 °C and then exposed to ultrasonication (Sonoplus, Bandelin, sonotrode MS73) three times, for 15 s each, at cycle 4 × 10% with the power set to 65%. The release of fatty acids was determined using the NEFA-HR (2) assay (Wako Fujifilm) as described previously (15).

Liposomes were generated by using PG or PC (Avanti Polar Lipids, Inc.). A total of 8 mM of the lipid was resuspended in 40 mM Tris HCl (pH = 7.5) with 10% glycerol and incubated for 1 h at 37 °C at 250 rpm. The lipid suspension was sonicated with following parameters: 3 × 15 s at cycle 4 and 65% intensity (Bandelin Sonopuls, sonotrode MS73). A total of 1 mL of 1 mM liposome suspension was incubated with 0.3 µM purified PlaB D203N, PlaB D203N Δβ9/10 for 30 min followed by ultracentrifugation at 38,800 rpm at 4 °C for 30 min. The supernatants were 10-fold concentrated (Amicon Ultra-4 10K Centrifugal Filter Devices, Merck), and the pellet was resuspended in 100 µL 40 mM Tris HCl (pH = 7.5). Detection of Strep-PlaB was performed via Western blot analysis (StrepMAB-Classic, horseradish peroxidase [HRP] conjugate, Iba).

L. pneumophila ΔplaB was grown over night in BYE to optical density (OD₆₀₀) ∼3. Bacterial cells were harvested and washed 2× with 40 mM Tris HCl (pH = 7.5). A total of 1 mL of the cells was incubated with 0.3 µM purified PlaB or PlaB Δβ9/10 for 10 min at 37 °C. Cells were washed 10× with 1 mL 40 mM Tris HCl. The cell pellet was lysed in 50 µL 40 mM Tris HCl (pH = 7.5) with 10 mg/mL lysozyme and Triton X-100 (0.1%) and was incubated at 37 °C for 30 min. After lysis, the volume of the sample was adjusted to 1 mL with Tris HCl. The final lysates and supernatants of all wash steps were used for Western blot analysis detecting Strep-PlaB (StrepMAB-Classic, HRP conjugate, Iba).

Separation of inner and OMs of E. coli and L. pneumophila strains was performed with changes according to Roy and Isberg (62). For E. coli, 100 mL of cell culture were induced at an OD₆₀₀ = 0.8 with 0.1 mM IPTG and incubated for 3 h at 37 °C and 250 rpm. For L. pneumophila, the induced culture was grown until late exponential phase. The culture was adjusted to OD₆₀₀ = 1. The pellet of 100 mL was resuspended in 1 mL cold fractionation buffer (40 mM Hepes, 0.1 mM ethylenediaminetetraacetic acid, 1× cOmplete Protease Inhibitor Mixture [Roche Sigma]) and the pellet of 1 mL lysed with Bug Buster (pellet sample) (Bug Buster Master Mix, Novagen Merck). For lysis, the cells were sonified for 3 min (30% amplitude, 0.4-s pulse, Bandelin Sonoplus HD2070). The supernatant (soluble components), obtained after centrifugation at 10,000 × g for 20 min at 4 °C, was ultracentrifuged at 100,000 × g for 1 h at 4 °C. The pellet (membrane fraction) was washed and resuspended in 1 mL cold fractionation buffer, and 2% Triton X-100 and 30 mM MgCl₂ were added (62, 63). After 5 min at room temperature, the pellets were resolved in an ultrasound bath (5 min, room temperature Bandelin Sonorex Super RK 103 H), and ultracentrifuged at 100,000 × g for 2 h at 4 °C. The supernatant (inner membrane fraction) was separated from the pellet (OM), which was dissolved in 1 mL fractionation buffer using an ultrasound bath for 5 min. Fractions were analyzed by means of Western blotting detecting Strep-PlaB (StrepMAB-Classic, HRP conjugate, Iba), the OM control proteins OmpA [α-OmpA was kindly provided by Nemani V. Prasadarao, Children's Hospital Los Angeles, Los Angeles, CA (64)] or MOMP (α-MOMP, MONOFLUO L. pneumophila IFA Test Kit, BioRad), the inner membrane control protein LepB [antibody kindly provided by Gunnar von Heinje, Stockholm University, Stockholm, Sweden (65)], and the cytosolic control proteins DnaK (DnaK monoclonal antibody, Enzo Life Sciences).

The E. coli BL21 and the L. pneumophila strains were grown to an OD₆₀₀ of 1.0 and expression was induced by the addition of 0.1 mM IPTG. At distinct optical densities, 1 mL samples were pelleted. The supernatants were concentrated 10 times with Amicon Ultra 0.5 mL Centrifugal Filters (−10 K, Merck) and boiled with 1× sodium dodecyl sulfate (SDS) loading buffer. The cell pellets were resuspended in 50 µL 40 mM Tris HCl (pH = 7.5) with lysozyme [10 mg/mL] and Triton X-100 [0.1% vol/vol] and incubated at 37 °C for 30 min. After lysis, the volume of the samples was adjusted to 1 mL with 40 mM Tris HCl (pH = 7.5) and sample aliquots for SDS–polyacrylamide gel electrophoresis (PAGE) and Western blotting analysis were prepared. Strep-PlaB and cytoplasmic DnaK in E. coli strains was detected by StrepMAB-Classic (HRP conjugate, Iba) and DnaK (DnaK monoclonal antibody, Enzo Life Sciences) antibodies. For detection of 4×HA-PlaB and cytoplasmic ICDH of L. pneumophila, α-HA.11 Epitope Tag (BioLegend) and α-ICDH [kindly provided by Ralph Isberg, Tufts University, Boston, MA (66)] antibodies were used.