The phosphatidic acid pathway enzyme PlsX plays both catalytic and channeling roles in bacterial phospholipid synthesis

PlsX is the first enzyme in the pathway that produces phosphatidic acid in Gram-positive bacteria. It makes acylphosphate from acyl-acyl carrier protein (acyl-ACP) and is also involved in coordinating phospholipid and fatty acid biosyntheses. PlsX is a peripheral membrane enzyme in Bacillus subtilis, but how it associates with the membrane remains largely unknown. In the present study, using fluorescence microscopy, liposome sedimentation, differential scanning calorimetry, and acyltransferase assays, we determined that PlsX binds directly to lipid bilayers and identified its membrane anchoring moiety, consisting of a hydrophobic loop located at the tip of two amphipathic dimerization helices. To establish the role of the membrane association of PlsX in acylphosphate synthesis and in the flux through the phosphatidic acid pathway, we then created mutations and gene fusions that prevent PlsX's interaction with the membrane. Interestingly, phospholipid synthesis was severely hampered in cells in which PlsX was detached from the membrane, and results from metabolic labeling indicated that these cells accumulated free fatty acids. Because the same mutations did not affect PlsX transacylase activity, we conclude that membrane association is required for the proper delivery of PlsX's product to PlsY, the next enzyme in the phosphatidic acid pathway. We conclude that PlsX plays a dual role in phospholipid synthesis, acting both as a catalyst and as a chaperone protein that mediates substrate channeling into the pathway.

PlsX is the first enzyme in the pathway that produces phosphatidic acid in Gram-positive bacteria. It makes acylphosphate from acyl-acyl carrier protein (acyl-ACP) and is also involved in coordinating phospholipid and fatty acid biosyntheses. PlsX is a peripheral membrane enzyme in Bacillus subtilis, but how it associates with the membrane remains largely unknown. In the present study, using fluorescence microscopy, liposome sedimentation, differential scanning calorimetry, and acyltransferase assays, we determined that PlsX binds directly to lipid bilayers and identified its membrane anchoring moiety, consisting of a hydrophobic loop located at the tip of two amphipathic dimerization helices. To establish the role of the membrane association of PlsX in acylphosphate synthesis and in the flux through the phosphatidic acid pathway, we then created mutations and gene fusions that prevent PlsX's interaction with the membrane. Interestingly, phospholipid synthesis was severely hampered in cells in which PlsX was detached from the membrane, and results from metabolic labeling indicated that these cells accumulated free fatty acids. Because the same mutations did not affect PlsX transacylase activity, we conclude that membrane association is required for the proper delivery of PlsX's product to PlsY, the next enzyme in the phosphatidic acid pathway. We conclude that PlsX plays a dual role in phospholipid synthesis, acting both as a catalyst and as a chaperone protein that mediates substrate channeling into the pathway.
Most bacteria synthesize phosphatidic acid (PA), 5 the central precursor of membrane phospholipids, using an unusual acylphosphate intermediate in a three-step pathway mediated by the PlsX, PlsY, and PlsC enzymes (1, 2) (Fig. 1A). PlsX is an acyl-ACP:phosphate transacylase that converts acyl-ACP made by FASII into acylphosphate (acyl-PO 4 ). Next, PlsY, a glycerol 3-phosphate (G3P) acyltransferase (GPAT), which uses exclusively acyl-PO 4 , transfers the acyl group of acyl-PO 4 to position 1 of G3P to produce lysophosphatidic acid (LPA). The last step of the pathway is catalyzed by PlsC, an LPA acyltransferase (AGPAT), which uses acyl-ACP to convert LPA into PA (1,3). PlsX and PlsY are found exclusively in bacteria and are essential in important human pathogens (1), therefore, these enzymes are considered promising targets for new antibacterial molecules.
cro ARTICLE from the cytoplasm was poorly understood until the recent determination of the crystal structures of PlsY (5) and PlsC (6). PlsY is a typical integral membrane protein with seven transmembrane segments delimiting a V-shaped active site that extends deeply into the bilayer. Access of its substrate acyl-PO 4 to the active site seems to occur from within the membrane, through a putative lateral gate between transmembrane helices 2 and 5, whereas G3P enters the active site via its cytoplasmic opening (5). The crystal structure of PlsC, on the other hand, showed that the protein lacks transmembrane segments and is instead anchored to the cytoplasmic leaflet of the bilayer via a novel motif comprised of two ␣ helices that lay parallel to the membrane (6). This mode of anchoring places the active site of PlsC above the bilayer in a way that it can simultaneously access LPA residing within the membrane and acyl-ACP from the cytoplasm (6). Despite these advances, topological information is still not available for the first reaction in the pathway, acyl-PO 4 formation by PlsX.
Based on sequence analysis and fractionation, PlsX was originally thought to be a soluble enzyme but subsequent work showed that it is a peripheral membrane protein (3,7,8). One of these publications reported that PlsX localization on membranes followed that of the bacterial cell division apparatus but our group has challenged this conclusion by showing that PlsX is evenly distributed on the membrane of proliferating cells, and that its localization was independent of cell division proteins (8). We also showed that PlsX association with the membrane was independent of PlsY, and that both inhibition of phospholipid synthesis and changes in membrane potential perturbed PlsX localization (8). These data suggested that PlsX associates directly with the phospholipid bilayer, but this possibility has not been tested experimentally.
Currently, two crystal structures of PlsX are available, those from Bacillus subtilis (9) and from the human pathogen Enterococcus faecalis (10). In both structures PlsX forms an S-shaped dimer. Dimerization is mainly mediated by hydrophobic contacts between a long amphipathic helix (␣9) and ␣10 in each monomer, which together form a tip that protrudes away from the core of the enzyme. This dimeric structure is further stabilized by additional hydrophobic interactions with the enzyme core and several intrasubunit salt bridges on the protein surface. The core domain of the enzyme comprises a long ␤-strand surrounded by ␣-helices, generating a deep groove where the active site is located (10). Both structures suggest that this tip might participate into the association of PlsX with the membrane, however, the membrane-binding mechanism and how membrane association allows PlsX to perform its transacylase reaction are largely unknown.
In this study, we report that PlsX binds directly to lipid bilayers and identified a loop rich in hydrophobic residues localized at the tip of PlsX, which plays a critical role in mediating protein-membrane interaction. We then exploited mutations and gene fusions in PlsX to assess the role of membrane association in acyl-PO 4 production and flux through the PA pathway. Mutations that severely affect the association of PlsX to the membrane do not eliminate its transacylase enzymatic activity. Strikingly, PA synthesis is severely hampered in these cells, probably because membrane association is required for the proper delivery of PlsX's product, acyl-PO 4 , to PlsY, its partner protein. Thus, PlsX plays a dual role in PA synthesis, acting both as a catalyst and as a chaperone protein that mediates substrate channeling into the pathway.

Identification of the membrane-interacting domain of PlsX
To determine how PlsX associates with membranes, we analyzed its structure (9, 10) and amino acid sequence searching for typical membrane-interacting domains. PlsX lacks terminal amphipathic helices like those mediating the association of peripheral proteins such as MinD and FtsA with the membrane (11,12), but we found a hydrophobic loop (containing 70% hydrophobic residues, H ϭ 0.648 and GRAVY value of 1.56, over 10 amino acids), which is solvent-exposed and next to a stretch of cationic amino acids (lysine residues) (Figs. 1B and 2A). This loop, which connects helices ␣9 and ␣10 and is disordered in the crystal structure, could play a role in targeting PlsX to the membrane (Fig. 1C). To test this possibility, we replaced some of the hydrophobic residues in the loop to glutamates and some of the basic amino acids in the adjacent ␣10 helix to alanine residues and investigated how these mutations affected PlsX's ability to interact with phospholipids in vitro. We expressed and purified each mutant with a C-terminal His 6 tag and assayed their association with liposomes made with B. subtilis lipids. As shown in Fig. 2B, PlsX-His 6 sedimented with liposomes but when the same experiment was repeated with mutants in the hydrophobic loop (PlsX L254E , PlsX L258E-A259E , and PlsX V262E ), little or no protein was detected in the pellet, indicating the importance of these residues for lipid binding. In contrast, PlsX mutants with substitutions in positive residues (PlsX K264A and PlsX K271A ) and in tyrosine 276 (PlsX Y276E ) of the ␣10 helix sedimented with liposomes (Fig. 2B).
We also investigated the interaction between PlsX and membranes by differential scanning calorimetry (DSC). Fig. 2C shows the heat capacity profiles of 1,2-dimiristoyl-sn-glycero-3-phospho-(1Ј-rac glycerol) (DMPG) large unilamellar vesicles (LUVs) mixed with PlsX at different lipid-to-protein (L/P) molar ratios. Binding of PlsX to DMPG LUVs decreased the enthalpy change (⌬H) associated with the lipid phase transition (Table 1), and made this transition broader and less cooperative without changing the transition melting temperature (T m ). DSC was also employed to examine the role of head group charge in the PlsX-membrane interaction. As illustrated in Fig.  S1, WT PlsX binds to LUVs made with another anionic phospholipid (phosphatidylserine, 1,2-dimyristoyl-sn-glycero-3phospho-L-serine (DMPS)), but does not bind to LUVs made with a zwitterionic lipid (phosphatidylcholine, 1,2-dimyristoylsn-glycero-3-phosphocholine (DMPC)). This result highlights the importance of electrostatic interactions for PlsX-membrane association.
We next probed our mutants with DSC and observed behaviors that were in complete agreement with the sedimentation experiments: both lysine residue mutants (PlsX K264A and PlsX K271A ) perturbed phase transition like the WT protein ( Fig.  2C and Table 1), whereas the double mutant PlsX L258E-A259E showed a partial effect compatible with its residual sedimenta-Catalytic and channeling roles of PlsX tion with liposomes. Interestingly, the PlsX L254E mutant had no effect on phase transition, confirming that this mutant has a severely compromised interaction with membranes ( Fig. 2C).
To confirm that the hydrophobic loop identified in vitro was indeed necessary for membrane localization of PlsX in vivo we tagged the PlsX L254E and PlsX K271A mutant proteins with GFP and determined their subcellular localization by fluorescence microscopy. We used a strategy in which the GFP-tagged mutants were integrated at the nonessential amyE locus and expressed from an IPTG-responsive promoter (P hyspank ) in a strain where endogenous PlsX could be depleted by growth in the absence of xylose (DS28 background). Thus, the localization of mutant GFP-PlsX proteins was evaluated without the interference of endogenous PlsX, which could potentially hetero-oligomerize with the mutant proteins. As shown in Fig.  2D, the mutant proteins localized as expected from their ability to bind to lipids in vitro: GFP-PlsX K271A was found in the membrane, like WT GFP-PlsX (8), whereas GFP-PlsX L254E exhibited a cytosolic localization. Altogether, our data indicate that the hydrophobic loop between helices ␣9 and ␣10 plays a prominent role in the association and localization of PlsX to the membrane, both in vitro and in vivo.

Dimerization of PlsX is important for membrane interaction
The crystal structures of B. subtilis and E. faecalis PlsX showed that the protein is a dimer maintained mostly by interactions involving helices ␣9 and ␣10, which flank the hydrophobic loop responsible for membrane insertion (9, 10) ( Fig.  1C). Helix ␣9, in particular, is a long amphipathic helix (AH) whose hydrophobic side makes many of the contacts that stabilize the PlsX dimer. To test the role of dimerization for membrane targeting we constructed three mutants: (i) PlsX L235E in which a conserved Leu in the middle of ␣9 was mutated to a negatively charged Glu residue; (ii) PlsX AH(ϩ2) in which two residues (Ala-Glu) were inserted between Ile 242 and Phe 243 to maintain the helicity but destroy the polar-apolar residue phasing of ␣9; and (iii) PlsX AH(ϩ3) in which the insertion of three residues (Glu-Glu-Leu) between Ile 242 and Phe 243 should maintain both the helicity and the amphipathic phasing of ␣9 (Fig.  3A). The three mutants were expressed and purified and their oligomerization state was analyzed by gel filtration alongside the WT protein (Fig. 3B). As expected from the crystal structures and previously shown for Streptococcus pneumoniae PlsX   Fig. S2). Significant changes in the elution pattern were observed for mutants PlsX AH(ϩ2) and PlsX L235E , which displayed retention times corresponding to the monomer (ϳ40 kDa), in addition to large aggregates (Fig. 3B), which probably arise because the mutant proteins can no longer dimerize. In contrast, mutant PlsX AH(ϩ3) , which bears a larger insertion than PlsX AH(ϩ2) but whose amphipathicity is maintained, behaved as a mixture of both monomeric and dimeric species (Fig. 3B).

Catalytic and channeling roles of PlsX
To examine whether dimerization of PlsX is relevant for membrane association, the above mutants were fused to GFP and their cellular distribution was examined, employing the same strategy used to study the localization of the hydrophobic loop mutants (expression of mutant GFP-PlsX fusions from an IPTG-inducible promoter in the absence of endogenous PlsX). As shown in Fig. 3C, the two mutants that no longer dimerized, PlsX L235E and PlsX AH(ϩ2) , had severely compromised membrane localization, as inferred by their cytoplasmic staining. In Table 1 Thermodynamic parameters associated to the DMPG phase transition The phase transition temperature (T m ), the calorimetric enthalpy change (⌬H), and the linewidth at half height (⌬T1 ⁄2 ) were obtained from analysis of the thermograms with MicroCal Origin software. Lipid-to-protein molar ratios (L/P) are indicated. Data shown are representative of two independent experiments.

Catalytic and channeling roles of PlsX
contrast, mutant PlsX AH(ϩ3) , which is partially a dimer in vitro (Fig. 3B), exhibited a mixed localization pattern, with both membrane and cytoplasmic staining (Fig. 3C). The importance of dimerization for membrane association was corroborated by in vitro experiments in which we observed that monomeric PlsX L235E no longer co-sedimented with liposomes, whereas the partially dimeric PlsX AH(ϩ3) still sedimented, albeit to a lesser extent than the WT protein (Fig. S3).

Membrane association is necessary for PlsX function in vivo
The availability of mutants with severe defects in membrane association provided an opportunity to investigate if PlsX's membrane localization was necessary for its functioning in vivo. The function of PlsX mutants was monitored by their ability to support growth in the absence of WT PlsX. We created merodiploid strains with mutant PlsX alleles controlled by IPTG in our PlsX depletion strain background and assayed growth by spotting dilutions on plates containing IPTG but no xylose. As shown in Fig. 4A, there was a clear correlation between membrane association and the ability to support growth: mutants that were affected in membrane association because of alterations in the hydrophobic loop (PlsX L254E ) or because of impaired dimerization (PlsX L235E , PlsX AH(ϩ2) ) supported little or no growth in the absence of endogenous PlsX, mutant PlsX AH(ϩ3) , which exhibited less affected membrane localization, promoted intermediate levels of growth, and mutant PlsX K271A , which associates with the membrane, was able to fully support growth (Fig. 4A). Physpank-gfp-plsX L235E ) were grown in liquid LB to logarithmic phase, serially diluted and spotted (4 l) onto LB ϩ 0.5 mM IPTG plates, with or without 0.4% xylose, and incubated at 37°C for 24 h. B, in vitro acyltransferase activity assay of PlsX wt and mutants. The assay detects the synthesis of acyl-ACP from 16:0-PO 4 and holo-ACP by the appearance of a slow migrating band in conformationally sensitive gel electrophoresis (15% PAGE, 2 M urea). Data shown are representative of three independent experiments. Note that the PlsX AH (2) sample was run in a different gel and thus is depicted separate from the other samples. , and insertion of two (AH(ϩ2)) and three residues (AH(ϩ3)) mutants are represented. Diagrams generated using the web server HELIQUEST (32) B, gel filtration chromatography assays. Purified PlsX wt (red) was a dimeric protein, whereas PlsX AH(ϩ2) (blue) and PlsX L235E (black) presented monomeric and aggregated states. PlsX AH(ϩ3) (green) present both dimeric and monomeric states. Data shown are representative of two independent experiments. C, fluorescence microscopy showing subcellular localization of GFP-tagged PlsX wt (strain DS69; Pxyl-plsX amyE::Physpank-gfp-plsX), PlsX AH(ϩ2) (strain DS34; Pxyl-plsX amyE::Physpank-gfp-plsX AH(ϩ2) ), PlsX AH(ϩ3) (strain DS48; Pxyl-plsX amyE::Physpank-gfp-plsX AH(ϩ2) ), and PlsX L235E (strain DS58; Pxyl-plsX amyE::Physpank-gfp-plsX L235E ) mutants. Images are representative of at least 100 cells analyzed for each condition in at least two different days. All images are at the same magnification. Scale bar ϭ 5 m.

Catalytic and channeling roles of PlsX
An important control to determine the role of PlsX localization in vivo was to rule out that the mutations affecting membrane association may have also affected catalysis by the enzyme. To study catalysis, we developed an in vitro acyltransferase assay measuring the reverse of the reaction normally executed by PlsX. We employed palmitoyl-PO 4 and holo-ACP as substrates and monitored the appearance of palmitoyl-ACP by conformation-sensitive gel electrophoresis. As shown in Fig.  4B, mutations that impaired dimerization (PlsX L235E , PlsX AH(ϩ2) ) abolished the production of palmitoyl-ACP by PlsX in vitro. This suggests that catalysis requires dimerization, a conclusion supported also by the fact that a mutation that rescues dimerization (PlsX AH(ϩ3) ) is catalytically active, even though it introduces a larger insertion in helix ␣9 than one that is inactive (PlsX AH(ϩ2) ). Importantly, however, a mutation that compromises membrane association without affecting dimerization (PlsX L254E ) did not affect catalysis in vitro (Fig.  4B). Therefore, the PlsX L254E mutation demonstrates that membrane binding and catalysis can be uncoupled in PlsX and we further characterized this mutation to understand why membrane localization is necessary for PlsX function in vivo.

Removal of PlsX from the membrane blocks phospholipid synthesis and leads to free fatty acid accumulation
The growth impairment of cells expressing PlsX L254E was an indication that the phospholipid synthesis pathway becomes blocked when PlsX cannot appropriately associate with the membrane. To test this possibility, we investigated if PlsX L254E growth could be rescued by exogenous fatty acids, as was shown for a plsX null mutant (13). In Gram-positive bacteria such as Bacillus and Staphylococcus, exogenous fatty acids can bypass the need for PlsX because they are phosphorylated directly by fatty acid kinase (Fig. 1A). Indeed, we found that supplementation of a strain expressing PlsX L254E with a mixture of branched-chain fatty acids rescued its growth as well as it rescued growth of a plsX null mutant (Fig. 5A), a strong indication that the growth defect of the PlsX L254E mutant was due to inhibition of phospholipid synthesis.

Catalytic and channeling roles of PlsX
We envisaged two possible scenarios to explain why phospholipid synthesis was compromised in cells expressing PlsX L254E : (i) the mutant protein is not catalytically active in vivo, or (ii) it is catalytically active but membrane association of PlsX is necessary for efficient delivery of acyl-PO 4 to PlsY to proceed with phospholipid synthesis. We reasoned that the two possibilities could be distinguished by comparing the de novo lipid synthesis profile of a PlsX L254E mutant with those of plsX Ϫ and plsY Ϫ mutants. If PlsX L254E is inactive in vivo it should exhibit a plsX Ϫ phenotype, but if the block in the pathway is at the level of product delivery, PlsX L254E should exhibit a plsY Ϫ phenotype. As reported previously by our group, a plsX Ϫ strain has a strong block in de novo lipid synthesis, presumably because the accumulation of acyl-ACP shuts down FASII (3). In contrast, de novo lipid synthesis is only partially inhibited in a plsY Ϫ strain, but this mutant accumulates free fatty acids, likely derived from the acyl-PO 4 produced by PlsX and which cannot be used further in the absence of PlsY (3). Therefore, we used [ 14 C]acetate labeling to analyze lipid synthesis in a strain expressing GFP-PlsX L254E , and compared it with a strain expressing a WT GFP-PlsX fusion, a conditional plsX mutant (plsX Ϫ ) and a conditional plsY mutant (plsY Ϫ ). As show in Fig.  5B, [ 14 C]acetate incorporation in a strain expressing GFP-PlsX L254E was partially inhibited, reaching ϳ30% of the WT levels, a value much higher than the incorporation observed in the conditional plsX mutant (7.0%) but similar to that observed in a conditional plsY mutant (Fig. 5B) (3). The inhibition of the GFP-PlsX L254E mutant was not due to a negative effect of the GFP fusion because lipid synthesis was normal in a strain expressing a WT GFP-PlsX fusion (Fig. 5B). In addition, TLC analysis of the newly synthesized lipids showed that the GFP-PlsX L254E mutant accumulated free fatty acids to a level comparable with the conditional plsY mutant (Fig. 5C). Thus, collectively, these data are most consistent with the hypothesis that GFP-PlsX L254E is catalytically active in vivo and that the main defect of this mutant protein is in channeling its product into the next step of the pathway.

Mistargeting as a way to test the role of membrane localization in PlsX function
Another way to test if membrane localization is necessary for PlsX function is to artificially target the protein to a different subcellular compartment without mutating the protein's amino acid sequence. We attempted to do that by fusing PlsX to HBsu, a histone-like protein that binds to the bacterial nucleoid with great affinity, such that normally all HBsu in the cell is nucleoid-associated (14). The high affinity of HBsu for the nucleoid and the fact that this structure is spatially segregated from the membrane suggested that fusing PlsX to HBsu might suffice to "peel" it from the membrane (Fig. 6A). However, when we imaged a strain expressing a triple GFP-PlsX-HBsu fusion we observed that the protein localized as foci or patches that overlapped with the nucleoid but that also reached the membrane, indicating that the fused protein manages to bind simultaneously to both structures (Fig. 6, B and C). In retrospect, this observation is not unexpected because it is known that DNA loops are frequently found in close proximity to the bacterial membrane (15,16).
Although the strategy of fusing PlsX to HBsu did not work to remove the WT protein from the membrane, we reasoned it could still be employed to further evaluate the importance of membrane association for PlsX function. Even though the PlsX L254E mutant displays a strongly reduced interaction with the membrane, it still supports residual growth when expressed in place of endogenous PlsX (Fig. 4A, compare spots of PlsX L254E and the parental depletion strain). Furthermore, a detailed biophysical analysis of the PlsX L254E mutant showed that it still retains the ability to bind superficially to membranes, probably via electrostatic interactions (33). Therefore, it is certainly possible that the cytoplasmic molecules of PlsX L254E will exhibit transient or superficial contact with the membrane sufficient to allow some transfer of acyl-PO 4 to the bilayer and maintain a low level of phospholipid synthesis. Alternatively, the residual growth may result from acyl-PO 4 produced in the cytoplasm diffusing and spontaneously inserting itself in the membrane. We reasoned we could distinguish between these possibilities by fusing PlsX L254E to HBsu to completely restrict access of the protein to the membrane. In fact, as shown in Fig.  6, B and C, fusing PlsX L254E to HBsu led to a dramatic change in its localization, which went from cytoplasmic to being exclusively associated with the nucleoid. We then asked if this protein, which now had no access to the membrane, was capable of supporting growth in the absence of endogenous PlsX. As shown in Fig. 6D, GFP-PlsX L254E -HBsu was slightly worse than GFP-PlsX L254E at complementing the lack of PlsX. The similar phenotypes of GFP-PlsX L254E and GFP-PlsX L254E -HBsu were also evident in the [ 14 C]acetate labeling experiments: both mutants exhibited comparable inhibition of lipid synthesis (Fig.  5B). However, the strain expressing GFP-PlsX L254E -HBsu accumulated 2-fold more free fatty acids (Fig. 5C), a sign that phospholipid synthesis is indeed more impaired when PlsX is completely prevented from accessing the membrane. The worse performance of GFP-PlsX L254E -HBsu cannot be explained as a general negative effect of HBsu, because fusion of WT PlsX to HBsu (GFP-PlsX-HBsu) had little if any impact on growth (Fig.  6D) or lipid synthesis (Fig. 5, B and C). We also carried out a control, which showed that all protein fusions were successfully expressed (Fig. S4). Therefore, these results reinforce our conclusion that PlsX membrane association is crucial for the efficient channeling of acyl-PO 4 into the phospholipid synthesis pathway.

Discussion
Peripheral proteins associate with the membrane in a wide variety of ways but two of the most common motifs are amphipathic helices and protruding hydrophobic loops (17). Here we showed that B. subtilis PlsX associates with the membrane via a hydrophobic loop comprising residues 254 -263 that projects from the end of its amphipathic dimerization helices. Typical loops have exposed hydrophobic residues at their tips, flanked by cationic residues that interact favorably with the negative head groups of lipids (18,19). In agreement with our prediction that the loop is the motif that targets PlsX to the membrane, mutations in hydrophobic residues (L254E, L258E/ A259E, and V262E) perturb to varying degrees membrane interaction. In contrast, we failed to disrupt membrane binding Catalytic and channeling roles of PlsX when single lysine residues (i.e. Lys-264 and Lys-271) of the adjacent ␣10 helix were mutated. This suggests that the energetic contribution of each individual positive residue for membrane binding is modest, but we cannot rule out the possibility that the ␣10 helix is not directly involved in membrane binding.
The existence of a hydrophobic loop flanked by positive residues in helix ␣10 is a feature conserved in all PlsX homologues (Fig. S5). In proteobacteria, this region is slightly longer (15 residues, instead of 10) and contains more aromatic and positive residues, but still has the features expected of a membranebinding hydrophobic loop. Hence, we expect that all PlsX proteins use similar motifs to associate with the membrane.
While this work was in preparation, Jiang and colleagues (20) described a new PlsX crystal structure in which the hydrophobic loop of one protomer is folded into a short amphipathic ␣ helix, suggesting that the region we have so far described as a hydrophobic loop may in fact change structure to become helical upon membrane interaction. This plausible model was supported by mutations that, similarly to ours, perturbed membrane interaction, but more detailed structural analysis in the presence of membranes will be necessary to ascertain the exact structure of PlsX. Irrespective of this structural detail, both works reinforce each other's conclusion that PlsX employs a stretch of hydrophobic and positively charged amino acids at the tip of its stalk as its membrane binding motif.
Another important observation of our work was that removal of PlsX from the membrane via mutation or the fusion to HBsu resulted in inhibition of phospholipid synthesis. This observation may be explained if membrane association was necessary for PlsX catalysis, but we find this very unlikely for two reasons: (i) PlsX catalysis can be detected in vitro in the absence of lipids or detergents (see Refs. 1 and 20, and this

Catalytic and channeling roles of PlsX
work), and (ii) we detect free fatty acid accumulation in cells expressing PlsX variants that do not localize to the membrane, confirming that the enzyme is catalytically active even when removed from the membrane. Why, then, would PlsX need to associate with the membrane to be fully functional? Being at the membrane interface makes sense for enzymes that act on substrates that are embedded in the bilayer, such as PlsC, which uses LPA, but this is not the case for PlsX whose substrates, acyl-ACP and P i , are accessible in the cytoplasm. Instead, we propose that PlsX association with the membrane is needed to channel acyl-PO 4 efficiently into the next step of the pathway, which is carried out by PlsY, an integral membrane protein that likely access acyl-PO 4 only after it has partitioned into the membrane (5) (Fig. 7).
Acylphosphates are less water soluble than acyl-ACP (2) and this may represent a barrier to its diffusion from the cytosolic compartment and into the membrane. Although amphipathic molecules like acyl-PO 4 , or even fatty acids, will spontaneously partition into the membrane from a water phase, the kinetics of the process may not be fast enough to support highly active pathways such as phospholipid synthesis (Fig. 7). In fact, the residual growth of strains containing PlsX variants that do not bind to the membrane (Figs. 4A and 6B) probably reflects the slow membrane partitioning of acyl-PO 4 that is being produced by PlsX molecules located in the cytoplasm. There are several examples of proteins that participate in cytoplasmic transport of fatty acids, whose diffusion is limited by their poor solubility (21). ACP itself can be thought of as an analogous acyl transporter, in the sense of avoiding solubility problems, because the acyl chain in an acyl-ACP molecule is tucked into a hydrophobic groove in the protein (22)(23)(24). Thus, it seems that acyl intermediates are not usually found freely surrounded by water inside cells, and PlsX would be the protein in charge of keeping acyl-PO 4 away from water as well. Finally, acylphosphates are very labile and the channeling (or chaperoning) role of PlsX may also be relevant to protect these molecules from hydrolysis, as pointed out by Li et al. (5). Consistent with this idea, we observe the accumulation of free fatty acids in strains in which PlsX is cytoplasmic and no longer capable of efficiently channeling acyl-PO 4 into the membrane (Fig. 5B).
The detailed mechanism of acyl-PO 4 channeling by PlsX remains to be defined, but our data allows us to speculate that it depends on the proper and stable insertion of PlsX in the membrane. This can be concluded from the biophysical analysis of the PlsX L254E mutant presented in the accompanying paper (33), which showed that this protein can interact with membranes only superficially. A similar conclusion is also supported by in vivo experiments, which showed that the phenotype of the PlsX L254E mutant is not much worse than the phenotype of a strain in which PlsX can never access the membrane (PlsX L254E -HBsu) (Fig. 6C). Together, these results mean that superficially accessing the membrane is not sufficient to ensure efficient transfer of acyl-PO 4 . Stable insertion in the membrane may be necessary to create a path shielded from solvent for acyl-PO 4 to leave the active site and insert into the bilayer. This path may also involve the physical association between PlsX and PlsY, as these proteins were shown to interact in two-hybrid experiments (25). Proper interaction between PlsX and PlsY may only happen if PlsX is stably inserted in the membrane.
In conclusion, we have identified a loop rich in hydrophobic residues localized at the tip of the stem of PlsX that plays a critical role in mediating protein-membrane interaction and used this information to advance our understanding of bacterial phospholipid biosynthesis. On the basis of our combined results, we propose that the proper membrane interaction and localization of PlsX plays a key topological role in the channeling of acyl substrates from the cytoplasm and into the phospholipid synthesis pathway. These results highlight the importance of spatial organization for the proper functioning of certain biochemical pathways within cells. In addition, they contribute to the field of antibacterial drug discovery, considering the importance of bacterial lipid synthesis as a target for new antimicrobial compounds.

Bacterial strains and growth conditions
All bacterial strains used in this study were listed in Table S1. Drug concentrations for selection were: erythromycin, 1 g ml Ϫ1 ; lincomycin 25 g ml Ϫ1 ; kanamycin, 50 g ml Ϫ1 ; chloramphenicol, 5 g ml Ϫ1 ; spectinomycin, 100 g ml Ϫ1 ; ampiciline 100 g ml Ϫ1. In general, B. subtilis and Escherichia coli were grown in Luria-Bertani (LB) broth medium or on plates at 37°C with aeration.

General molecular techniques
Molecular cloning techniques were performed as described (26). DNA fragments were obtained by PCR using Phusion DNA polymerase (New England Biolabs). Oligonucleotide primers were purchased from Exxtend SRL (Campinas, SP, Brazil) and Thermo Fisher Scientific (São Paulo, SP, Brazil) and are listed in Table S2. PCR products of the expected sizes were purified from gels using an GeneJET TM Gel Extraction Kit (Fermentas), ligated into the pTZ57R/T cloning vector (Fermentas) and transformed into E. coli DH␣ (Table S1). Plasmid DNA was prepared using a GeneJET Plasmid Miniprep kit (Thermo Sci- Figure 7. PlsX membrane association is crucial for acyl-PO 4 channeling to PlsY. Membrane association creates a path shielded from solvent for the transfer of acyl-PO 4 to the bilayer. Channeling may also involve transfer directly to PlsY, via protein-protein interaction. PlsX mutants that do not associate with the membrane still produce acyl-PO 4 but this cannot reach PlsY and is hydrolyzed to free fatty acids in the cytosol. Catalytic and channeling roles of PlsX entific). Sequencing was carried out using the Big Dye terminator cycle sequencing kit (Applied Biosystems) by the sequencing service of the Departamento de Bioquímica, IQ-USP (SP, Brazil). Transformation of B. subtilis was carried out by the method of Dubnau and Davidoff-Abelson (27). The amy phenotype was assayed for colonies grown for 24 h in LB-starch plates by flooding the plates with 1% I 2 , KI solution.

Plasmid and strain construction
All plasmids used in this study are listed in Table S1. The plsX gene from B. subtilis was amplified and inserted into pET24b (T7 promotor) within NdeI and XhoI sites to generate pET24-plsX, which expressed BsPlsX with a C-terminal His 6 tag. Sitedirected mutantagenesis was performed using Phusion Sitedirected Mutagenesis Kit (Thermo Fisher Scientific). GFP fusion constructs were designed for fluorescence microscopy. Construction of GFP-PlsX and GFP-PlsX-HBsu into pDR111 vector containing the double or triple fusion within NheI and SphI sites to generate pDR111-gfp-plsX and pDR111-gfp-plsXhbsu vectors to integrate in amyE locus of B. subtilis DS28 strains.

Expression and purification of BsPlsX
Vectors pET24b-plsX were used to transform E. coli BL21 (DE3) pRIL-competent cells for protein expression. BsPlsX WT and variant mutants were induced with 0.8 mM IPTG for 4 h at 37°C. Cells were collected by centrifugation (6000 rpm, 4°C, 15 min), and cell pellets were lysed with a sonicator. Soluble proteins were applied to a Ni 2ϩ -agarose column and washed with buffer: 50 mM Tris-HCl (pH 7.5), 300 mM NaCl, and 50 mM imidazole). His-tagged proteins were eluted with the same buffer containing 300 mM imidazole. Purified proteins were applied to a P10 desalting column and eluted with buffer A without imidazole. The proteins were stored at Ϫ80°C.

Gel filtration assay
The PlsX WT protein and variant mutants were purified by Ni 2ϩ column and then were subjected to gel filtration chromatography using Superdex 200 Increase 10/300 GL resin in a HPLC column (10.0 ϫ 300 mm) and employing the AKTA pure system (GE Healthcare). Proteins were eluted with buffer 50 mM Tris-HCl (pH 7.5), 300 mM NaCl at a flow rate of 0.5 ml/min and elution was monitored at 280 nm.

Fluorescence microscopy
For fluorescence microscopy, overnight precultures of B. subtilis were grown in LB medium containing antibiotics diluted in fresh medium without antibiotics to an optical density at 600 nm of 0.05 and grown at 37°C. At appropriate time intervals, 5 l of living cells was applied directly on glass slides on a thin film of 1.5% LB-agarose and examined using a ϫ100 oil immersion objective on a Nikon Eclipse Ti microscope, equipped with GFP BrightLineR Set (Semrock). Images were acquired on a system with the Nikon digital camera (Nikon, Tokyo, Japan). All images were captured using NIS software version 3.07 (Nikon) and fluorescence signals were analyzed using ImageJ software.

Synthesis of palmitoyl-PO 4
The 16:0-PO 4 was synthesized by a new mechanochemical method for the synthesis of acylphosphates. 6 Palmitoyl chloride (370 l, 1,2 mmol) was added to a mixture of tripotassium phosphate (K 3 PO 4 , 297 mg, 1.4 mmol) and 10 mol % tetra-nbutylammonium bromide (Bu 4 NBr, 40 mg), followed by trituration. Trituration was continued until the reaction was completed as evident from the formation of a solid (about 5 min). The resulting mixture was filtered in a celite pad, and the retained solid was washed with chloroform (3 ϫ 10 ml). The chloroform-soluble material was dried under vacuum to afford 400 mg of 16:0-PO 4 as a white powder (50% yield). 1

PlsX activity assay
The acyltransferase activity was determined according to the procedures described by Lu et al. (1). The reaction for acyl chain transfer from 16:0-PO 4 to ACP contained 0.1 M HEPES (pH 7.4), 0.05% Tween 20, 2 mM MgCl 2 , 40 M 16:0-PO 4 , and 40 M ACP. Reaction was started by adding BsPlsX (5-30 M) and was incubated at 37°C for 1 h with agitation. Samples were analyzed by conformationally sensitive gel electrophoresis in 15% polyacrylamide gels containing 2 M urea (34,35) and staining with Coomassie Brillant Blue.

Western blotting
Samples were prepared as described Weart et al. (28). Briefly, harvested cells were resuspended in 50 mM Tris-HCl (pH 7.5), 300 mM NaCl, 1 mM phenylmethylsulfonyl fluoride, 2 mg ml Ϫ1 of lysozyme. Cells were incubated at 37°C for 30 min and lysates were normalized by OD 600 at cell harvest, resuspended in sample buffer, and subjected to SDS-PAGE. GFP-PlsX was detected using the polyclonal rabbit anti-GFP antibody (29). FtsZ was detected using the polyclonal anti-FtsZ antibody (laboratory stock).

Liposome sedimentation assay
Initially, total phospholipids from B. subtilis PY79 were extracted with the Bligh and Dyer method (30). Liposomes were made by resuspension of the lipid film in 50 mM Tris-HCl (pH 7.5), 300 mM NaCl. For sedimentation assays, protein and liposomes were mixed at different concentrations in the assay buffer to a final volume of 150 l. After incubation at room temperature for 30 min, samples were centrifuged at 65,000 rpm for 15 min in a Beckman TLA100 rotor. Supernatants were removed immediately and pellets were resuspended in an equal volume of buffer. Samples from supernatant and pellet were subjected to 12% SDS-PAGE and proteins were detected by Coomassie Blue staining phospholipid vesicles.

DSC experiments
The effects of PlsX on lipid bilayer membranes composed of DMPG, DMPC, and DMPS were analyzed by monitoring the 6 Marques Netto, unpublished data.

Catalytic and channeling roles of PlsX
thermotropic phase behavior of lipid and lipid-protein samples using DSC. The thermograms were recorded in a VP-DSC MicroCal MicroCalorimeter (Microcal, Northampton, MA) using a heating rate of 23.4°C/h. The analyses of the heat capacity curves were performed using Microcal Origin software (OriginLab Corporation, Northampton, MA). Lipid vesicles were prepared as follows: aliquots of lipid stock solutions in chloroform/methanol, 1:1 (v/v), were added into a glass tube, dried under a N 2 flow, and ultracentrifuged under vacuum for 5 h to remove traces of organic solvent. The resulting lipid film was hydrated in 50 mM Tris-HCl buffer (pH 7.5), 300 mM NaCl at a temperature above the main phase transition of the lipid for 2 h and submitted to six freeze-thaw cycles. LUV were obtained by extruding the preformed multilamellar dispersions using polycarbonate filters with a pore size of 100 nm (Nuclepore Corp., Cambridge, CA). An aliquot of the protein stock solution was added into the LUVs, gently vortexed for few seconds, and kept at either 37°C (DMPC and DMPG) or 40°C (DMPS) for 1 h prior to the DSC experiment. The phospholipids were purchased from Avanti Polar Lipids, Inc. (Alabaster, AL) and used without further purification.