Branched-chain amino acid metabolism controls membrane phospholipid structure in Staphylococcus aureus

Branched-chain amino acids (primarily isoleucine) are important regulators of virulence and are converted to precursor molecules used to initiate fatty acid synthesis in Staphylococcus aureus. Defining how bacteria control their membrane phospholipid composition is key to understanding their adaptation to different environments. Here, we used mass tracing experiments to show that extracellular isoleucine is preferentially metabolized by the branched-chain ketoacid dehydrogenase complex, in contrast to valine, which is not efficiently converted to isobutyryl-CoA. This selectivity creates a ratio of anteiso:iso C5-CoAs that matches the anteiso:iso ratio in membrane phospholipids, indicating indiscriminate utilization of these precursors by the initiation condensing enzyme FabH. Lipidomics analysis showed that removal of isoleucine and leucine from the medium led to the replacement of phospholipid molecular species containing anteiso/iso 17- and 19-carbon fatty acids with 18- and 20-carbon straight-chain fatty acids. This compositional change is driven by an increase in the acetyl-CoA:C5-CoA ratio, enhancing the utilization of acetyl-CoA by FabH. The acyl carrier protein (ACP) pool normally consists of odd carbon acyl-ACP intermediates, but when branched-chain amino acids are absent from the environment, there was a large increase in even carbon acyl-ACP pathway intermediates. The high substrate selectivity of PlsC ensures that, in the presence or the absence of extracellular Ile/Leu, the 2-position is occupied by a branched-chain 15-carbon fatty acid. These metabolomic measurements show how the metabolism of isoleucine and leucine, rather than the selectivity of FabH, control the structure of membrane phospholipids.

Branched-chain amino acids (primarily isoleucine) are important regulators of virulence and are converted to precursor molecules used to initiate fatty acid synthesis in Staphylococcus aureus. Defining how bacteria control their membrane phospholipid composition is key to understanding their adaptation to different environments. Here, we used mass tracing experiments to show that extracellular isoleucine is preferentially metabolized by the branched-chain ketoacid dehydrogenase complex, in contrast to valine, which is not efficiently converted to isobutyryl-CoA. This selectivity creates a ratio of anteiso:iso C 5 -CoAs that matches the anteiso:iso ratio in membrane phospholipids, indicating indiscriminate utilization of these precursors by the initiation condensing enzyme FabH. Lipidomics analysis showed that removal of isoleucine and leucine from the medium led to the replacement of phospholipid molecular species containing anteiso/iso 17-and 19-carbon fatty acids with 18-and 20-carbon straight-chain fatty acids. This compositional change is driven by an increase in the acetyl-CoA:C 5 -CoA ratio, enhancing the utilization of acetyl-CoA by FabH. The acyl carrier protein (ACP) pool normally consists of odd carbon acyl-ACP intermediates, but when branched-chain amino acids are absent from the environment, there was a large increase in even carbon acyl-ACP pathway intermediates. The high substrate selectivity of PlsC ensures that, in the presence or the absence of extracellular Ile/Leu, the 2-position is occupied by a branched-chain 15-carbon fatty acid. These metabolomic measurements show how the metabolism of isoleucine and leucine, rather than the selectivity of FabH, control the structure of membrane phospholipids.
Fatty acid synthesis is a vital aspect of bacterial metabolism and carried out by a series of reaction steps catalyzed by individual proteins known as the type II fatty acid synthase (FASII) (1). Fatty acids are the major building blocks of the phospholipids, and the biophysical properties of cellular membranes are controlled in large part by the composition of the acyl chains (2). Bacteria alter the membrane fatty acid composition to modify membrane fluidity in response to the environment using two different biochemical schemes (Fig. 1). In model organisms like Escherichia coli and Streptococcus pneumoniae, membrane fluidity is altered by varying the saturated:unsaturated fatty acid ratio at a branch point within FASII (Fig. 1A) The enzymatic mechanisms and transcription factors that regulate unsaturated fatty acid synthesis are known in these systems (2)(3)(4)(5)(6)(7)(8)(9)(10)(11). Unsaturated fatty acid synthesis arises, and the expression levels/activities of 3-hydroxyacyl acyl carrier protein (ACP) dehydratase/ isomerase (FabA) and 3-ketoacyl-ACP synthase I (FabB) determine the amount of unsaturated fatty acid produced (12). Many Gram-positive bacteria do not produce any unsaturated fatty acid, and the ratio of branched-chain to straight-chain fatty acids is a major determinant of membrane fluidity (2,(13)(14)(15)(16)(17). Experiments with model membranes show that phospholipids with branched-chain fatty acids are more fluid than their straight-chain homologs resulting in reduced lipid bilayer thickness and an increase in membrane bilayer fluidity (18,19). Fatty acids with anteiso branching are more effective at fluidizing the membrane than fatty acid with iso branching. There is a correlation between an increase in the proportion of anteiso fatty acids as an adaptive response to reduced temperatures that is especially important for the survival of cold tolerant pathogens like Listeria monocytogenes (16,20,21). Unlike unsaturated fatty acid synthesis, the branched chain is introduced at the FabH step, the initial condensation reaction of FASII (Fig. 1B). Finally, some bacteria, like Bacillus and Pseudomonas, possess oxidative desaturases that introduce double bonds into existing fatty acids (22,23).
Branched-chain amino acids (Ile, Leu, and Val) are precursors to the acyl-CoA primers used by FabH to initiate FASII in the Gram-positive pathogen Staphylococcus aureus. The major fatty acids are derived from Ile based on the predominance of anteiso-branched-chain fatty acids present in the phospholipids (24). Extracellular amino acids are diverted to FASII by their conversion to their respective ketoacids by the IlvE transaminase (13,20,25). The ketoacids are decarboxylated by the branched-chain ketoacid dehydrogenase (Bkd) complex to form the respective acyl-CoAs (26). Val gives rise to isobutyryl-CoA (iC 4 -CoA), Leu is converted to isovaleryl-CoA (iC 5 -CoA), and Ile becomes 2-methyl-butyryl-CoA (aC 5 -CoA). Initiation with an amino acid-derived acyl-CoA results in branched-chain fatty acid synthesis, whereas initiation with acetyl-CoA (C 2 -CoA) gives rise to straight-chain fatty acids. In S. aureus, there is a single 3-ketoacyl-ACP synthase III (FabH) that draws from the short-chain acyl-CoA primer pool to initiate FASII (27,28). Purified FabH has a distinct preference for C 5 -CoA compared with C 2 -CoA but does not have a high selectivity for either iC 5 -CoA or aC 5 -CoA (13,27,28), making it difficult to understand how FabH can account for the preferential use of aC 5 -CoA to prime FASII or the modulation of fatty acid composition in response to environmental conditions. Changes in media composition and amino acid transporter activities (14,15) significantly alter fatty acid composition and are consistent with control being exerted by the supply rather than the utilization of FabH substrates. How amino acid metabolism is connected to FASII has not been adequately explored.
The goal of this work is to determine how branched-chain amino acid metabolism impacts the acyl-CoA and acyl-ACP intermediate pool compositions and membrane phospholipid structure in S. aureus. Isotopic tracer experiments with masstagged branched-chain amino acids show that the Bkd pathway produced a FabH precursor pool that reflects the anteiso:iso ratio found in phospholipids. Val metabolism to iC 4 -CoA is minimal. The removal of Ile elevates phosphatidylglycerol (PG) molecular species containing 18-and 20carbon straight-chain fatty acids. The ACP pool normally consists of odd carbon acyl-ACP and hydroxyacyl-ACP intermediates. Ile limitation decreases the 2-methylbutyryl-CoA pool and increases the level of even carbon acyl-ACP FASII intermediates initiated from the utilization of C 2 -CoA. These metabolomic measurements show how the nutritional state with respect to Ile and Leu regulates the synthesis of straightchain fatty acid for the 1-position acyl chain, whereas 15-carbon branched fatty acids remain the dominant 2position acyl chain.

Branched-chain amino acid utilization
The impact of branched-chain amino acids of the intermediate metabolite pools that support FASII was examined in wildtype S. aureus strain AH1263 (29) grown in defined medium as described under the Experimental procedures section. The acyl-CoA precursors available to FabH in S. aureus were analyzed by LC-MS/MS of the extracted CoA thioester pool, and the CoA thioester content was normalized by the inclusion of an [ 13 C 2 ] C 2 -CoA internal standard in each replicate. A representative scan of S. aureus grown in defined media shows that C 2 -CoA is the major CoA thioester ( Fig. 2A). The iC 5 -CoA and aC 5 -CoA species are resolved by chromatography and are the next most abundant thioesters. The aC 5 -CoA:iC 5 -CoA ratio was 3.5 illustrating that the precursor for anteiso fatty acids is the most abundant branched-chain primer available. CoA is a significant component, and low levels of malonyl-CoA and iC 4 -CoA are detected. Butyryl-CoA and propionyl-CoA are not detected when grown in defined growth medium.
The utilization of amino acids for membrane synthesis was traced using defined growth media with Ile, Leu, and Val substituted with [ 13 C 6 ]Ile, [d 3 ]Leu, and [ 13 C 5 , 15 N]Val to introduce specific mass tags that allow the origin of each fatty acid produced to be traced to its amino acid source. aC 5 -CoA derived from [ 13 C 6 ]Ile will have a +5 mass, iC 5 -CoA derived Figure 1. Two different schemes used by bacteria to control fatty acid structure and membrane biophysical properties. A, many bacteria have FASII elongation systems that produce saturated fatty acid (SFA) and unsaturated fatty acid (UFA) straight-chain fatty acids. These organisms control their membrane biophysical properties by varying the proportion of UFA at a branch point in the middle of the elongation cycle. The genetic and biochemical mechanisms that regulate branch points within FASII are best known in Escherichia coli (1, 2), which is diagrammed here. However, organisms that make straight-chain fatty acids and use the PlsX/Y/C pathway, like Streptococcus pneumoniae, also control composition using a genetically controlled branch point in the middle of FASII (10). B, many bacteria do not make UFA and instead produce predominantly iso and anteiso branched-chain fatty acids. These organisms alter the proportions of branched-chain fatty acids to modify their membrane biophysical properties (1,2). The proportion of straight iso and anteiso fatty acids are determined at the initiation step (FabH). In both schemes, the activities of the elongation condensing enzymes and acyltransferases control fatty acid chain length. FASII, type II fatty acid synthase. from [d 3 ]Leu will have a +3 mass, and iC 4 -CoA derived from [ 13 C 5 , 15 N]Val will have a +4 mass. Wildtype strain AH1263 was grown to midlog phase in defined medium, and switched to media containing the heavy amino acids and samples removed at 10, 20, and 40 min and analyzed by LC-MS/MS. Samples obtained at the 10-min time point show that (+5)C 5 -CoA derived from [ 13 C 6 ]Ile is the major species with (+3)C 5 -CoA derived from [d 3 ]Leu being less abundant (Fig. 2B). The ratio of (+5)C 5 -CoA:(+3)C 5 -CoA was 3.5. Approximately 10% of the C 5 -CoA pool was unlabeled corresponding to the contribution of 2-methylbutyrate and isovalerate arising from de novo Ile and Leu biosynthesis. We only detected a trace of (+4)iC 4 -CoA in our analysis consistent with the low abundance of iC 4 -CoA in the CoA thioester pool ( Fig. 2A). The contributions of Ile and Leu to the C 5 -CoA pool remained constant over the 40-min labeling period showing that the system reached a steady state within 10 min. These data show that growth in rich media typically used in the laboratory (like LB), the major C 5 -CoA precursor available to FabH is derived from extracellular Ile. Because the concentrations of Ile and Leu in the defined medium are both 50 μg/ml, these data mean that extracellular Ile is selectively metabolized by the Bkd pathway compared with Leu. Val is not a significant fatty acid precursor.
The flow of branched-chain amino acids into membrane phospholipids was assessed by determining the isotopic distribution in the PG molecular species over the time course. The correspondence between the MS peaks and the PG structures is shown in Figure S1. A sample mass spectrum obtained at the 20-min time point illustrates the primary data and how it was analyzed (Fig. 3A). The even and odd numbered PG molecular species had distinctly different labeling patterns. For example, three new peaks (+5, +8, and +10) appeared in the 32:0-PG (17:0/15:0) peak cluster. The (+5)32:0-PG peak corresponded to one anteiso-fatty acid derived from [ 13 C 6 ]Ile combined with another branched-chain fatty acid from the unlabeled C 5 -CoA precursor pool. Analysis of the fragmentation pattern of this peak showed that it was predominately 17:0/(+5)a15:0, although (+5)a17:0/15:0-PG was also present (Fig. 3B). The (+8)32:0-PG peak corresponded to one anteiso-fatty acid arising from Ile combined with an iso-fatty acid arising from Leu. Fragmentation of this peak showed that (+3)i17:0/(+5) a15:0-PG was the major species (Fig. 3C). The most abundant new peak in the 32:0-PG cluster had a +10 mass (m/z = 731), meaning that it contained two anteiso-fatty acids derived from Ile. Fragmentation of this peak confirmed it was composed of (+5)a17:0/(+5)a15:0-PG (Fig. 3D). This same ratio of +5, +8, and +10 molecular species was found in all even numbered PG species (30:0, 32:0, and 34:0) (Fig. 3A). Only a +5 peak was detected in the odd-numbered molecular species (29:0, 31:0, 33:0, and 35:0) corresponding to one unlabeled even-numbered fatty acid paired with an a15:0 derived from [ 13 C 6 ]Ile. Fragmentation of the (+5)33:0-PG confirmed that it was 18:0/(+5) a15:0-PG (Fig. 3E). We did not detect +4 peaks in the oddnumbered PG molecular species, meaning that Val is not the precursor to the even-carbon acyl chains. Rather, these evennumber fatty acids are not labeled indicating they arose from C 2 -CoA, the only available even-chain primer in the CoA thioester pool ( Fig. 2A). We did not detect +3 peaks in the evennumbered PG molecular species, meaning that the straightchain fatty acids are paired with a15:0 derived from Ile and not paired with i15:0 from Leu. The >2.2-fold higher signal of 15:0 in the fragmentation patterns of all PG molecular species indicates that 15:0 occupies the 2-position (30-32) consistent with previous analytical data showing that 15:0 almost exclusively occupies the 2-position of S. aureus phospholipids (24,33,34). The percentage of each heavy molecular species calculated from triplicate biological replicates illustrates the linear flow of amino acids into each molecular species over the time course (Fig. 3, F-H).

Isoleucine limitation and PG structure
The growth medium has a significant impact on the fatty acid composition of S. aureus (14,35), and striking changes are observed in the membrane PG molecular species in S. aureus in infection models (36,37). In the thigh infection model, even-chain fatty acids are paired with 15:0 predominated, and this effect could be created in the laboratory by removing Ile from the growth medium (36). However, these experiments did not address whether even carbon fatty acids were straight chain or iso branched (Val derived) or whether the odd carbon fatty acids are iso (Leu derived) or anteiso (Ile derived). Our baseline experiments validated these prior results (36) showing that the removal of Ile from the defined growth media resulted in a marked increase in PG molecular species containing evenchain fatty acid (Fig. S2). At Ile concentrations ≥40 μg/ml, even carbon molecular species predominated, whereas at Ile concentrations ≤10 μg/ml, odd and even carbon PG molecular species were equally abundant (Fig. 4A). Each of the odd PG molecular species decreased, and each of the even carbon species increased as Ile increased in the medium (Fig. S3).
Isotopic labeling was performed to determine how the absence of Ile from the medium alters the incorporation of Leu and Val into membrane phospholipids. Strain AH1263 was labeled with [d 3 ]Leu and [ 13 C 5 , 15 N]Val in the presence and absence of unlabeled Ile. Mass spectra in the presence of Ile showed detectable incorporation of Leu (+3) into all even carbon number PG molecular species. Val incorporation (+4) was only detected as a minor component of the 31:0-PG cluster (Fig. 4B). In the absence of Ile, the incorporation of Leu, and to a lesser extent Val, into PG was significantly enhanced (Fig. 4C). The even carbon molecular species (30:0, 32:0, and 34:0) had distinct +3 peaks indicating pairing of an iso fatty acid with an unlabeled iso or anteiso fatty acid. These even carbon number molecular species also had significant +6 peaks showing that iso fatty acids derived from Leu are incorporated into both positions of the glycerol backbone. The The individual molecular species are labeled above the brackets with the isotopically labeled peaks color coded according to their increased mass. Pre-existing and unlabeled peaks are black, the +5 peaks derived from [ 13 C 6 ]Ile are red, the +8 peaks derived from [ 13 C 6 ]Ile and [d 3 ]Leu are green, and the +10 peaks derived from [ 13 C 6 ]Ile are blue. The individual peaks were fragmented to determine the fatty acid composition. B, acyl chain composition of the (+5)32:0-PG peak. C, acyl chain composition of (+8)32:0-PG peak. D, acyl chain composition of (+10)32:0-PG peak. E, acyl chain composition of (+5)33:0-PG peak. Rates of Ile, Leu, and Val incorporation into the individual PG molecular species were obtained from three biological replicates, and the percentage contribution of each of the three amino acids to the total ion current in three of the prominent PG molecular species was calculated using LipidView software (Sciex). 29:0 and 31:0 PG species had detectable +4 peaks indicating the corporation of Val into these species as an even-carbon acyl chain. These Val-derived fatty acids are also paired with an iso fatty acid derived from Leu giving rise to a +7 peak in both cases. The 33:0 and 35:0 PG molecular species that are elevated in the absence of Ile (Fig. S2, C and D) have only +3 counterparts, meaning that these molecular species consist of an even-chain fatty acid that is not derived from amino acid metabolism and an iso fatty acid derived from Leu. These data illustrate the large impact of Ile availability on membrane phospholipid structure.

Acyl-CoA pool composition controls fatty acid structure
The impact of each of the four extracellular amino acid supplementation conditions on the abundance of acyl-CoA primers was analyzed by LC-MS/MS using three biological replicates (Fig. 6A). The removal of Leu from the medium reduces iC 5 -CoA and changes the aC 5 -CoA:iC 5 -CoA ratio from 3.5 to 27.3. The removal of Ile causes a marked reduction in aC 5 -CoA coupled with a large increase in iC 5 -CoA leading to a aC 5 -CoA:iC 5 -CoA ratio of 0.012. In the absence of both Ile and Leu, the overall concentration of branched-chain precursors is lower, but the aC 5 -CoA:iC 5 -CoA ratio of 3.4 is the same as in defined media. This result shows that the Ile is preferentially metabolized by the Bkd pathway in the absence of extracellular amino acids ruling out amino acid transporters as the selectivity filter. C 3 -CoA is not detected in defined media, but it contributes to the CoA thioester pool in media lacking Ile. The removal of Ile also results in an increase in the intracellular concentration of C 2 -CoA (Fig. 6A).
Gas chromatography reveals the close correlation between the acyl-CoA primer pool and fatty acid compositions (Fig. 6B  and Table S1). The fatty acid composition is dominated by anteiso fatty acid in defined medium, and Leu removal causes an increase in anteiso-odd fatty acids and a decrease in iso-odd acyl chains. The removal of Ile from the medium resulted in a large increase in iso fatty acids and a reduction in anteiso acyl chains. Removal of both Leu and Ile from the medium returned the anteiso:iso ratio in branched-chain fatty acids to that of media containing Ile and Leu, but the overall abundance of odd-chain fatty acids is lower. There is also an increase in iso-even-chain fatty acids (iso-14:0 and iso-16:0) derived from Val under conditions where extracellular Ile is absent (Fig. 6B and Table S1). Also, odd straight-chain fatty acids are found in cells grown without Ile (Table S1) that arise from FabH utilization of the propionyl-CoA pool that is detected in the absence of Ile. However, the largest increases arising from Ile removal are in even-numbered fatty acids that are not derived from amino acid metabolism (Fig. 6B). This correlates with a change in the C 2 -CoA:C 5 -CoA ratio from 0.86 in defined medium to 5.7 in the absence of extracellular Ile and Leu (Fig. 6A). These straight-chain fatty acids are primarily 18:0 and 20:0 (Table S1), leading to an increase in 33:0-PG and 35:0-PG and an increase in the average carbon number of the PG molecular species. The close link between the acyl-CoA abundance and the fatty acid compositions is evident comparing Figure 6, A and B. In the absence of extracellular Ile, S. aureus compensates by switching to the synthesis of fatty acids derived from C 2 -CoA that are incorporated into the 1-position of PG. These data suggest that the composition of the acyl-CoA primer pool is the major determinant of fatty acid composition and not the selectivity of FabH.

Impact of amino acids on the acyl-ACP pool
The removal of Ile from the medium triggers the synthesis of even-chain fatty acids not derived from branched-chain amino acids. We analyzed the levels of FASII intermediates in S. aureus to determine the rate-limiting steps in FASII and Ile-dependent alteration in the levels of pathway intermediates. We used a proteomics workflow adapted from Nam et al. (38) to prepare acyl-ACP fragments and performed four separate LC-MS/MS experiments on each biological replicate to profile all potential 3-ketoacyl-ACP, trans-2-enoyl-ACP (tC x -ACP), 3-hydroxyacyl-ACP (hC x -ACP), and acyl-ACP (C x -ACP) intermediates. [ 13 C 2 ]Acetyl-ACP was spiked into the samples at the beginning of the workflow and used as the internal standard to normalize the levels of intermediates in each biological replicate. The raw ion current data from representative analytic experiments to determine the acyl-ACP levels in S. aureus grown in defined medium is shown to illustrate the primary data (Fig. 7A). We did not detect a pool of ketoacyl-ACP. Control experiments using FabH to generate 3-ketoC 4 -ACP show that this intermediate is recovered by our workflow suggesting that these intermediates are rapidly metabolized by FabG in vivo. There was a series of odd-chain hC 7+2n -ACP detected (Fig. 7A). The relative abundance of these hC 7+2n -ACP intermediates was 10-fold lower than the abundance of the corresponding acyl-ACPs. tC 7 -ACP was the major trans intermediate detected with levels of other trans intermediates being low or below the level of detection (Fig. 7A). The relative abundance of tC 7 -ACP in the pathway suggests that metabolism of the first trans intermediate is the slowest step catalyzed by FabI in the pathway. Odd-carbon acyl-ACPs are the major pathway intermediates detected (Fig. 7A) illustrating the preferential use of branched-chain FabH substrates when Ile and Leu are present. Nonesterified ACP (ACPSH) is a major pool component with malonyl-ACP and acetyl-ACP also detected. There are also low signals for even-chain acyl-ACP detected in cells grown in defined medium (Fig. 7A). The detection of these even-chain acyl-ACPs in the pathway is consistent with the essential function of FASII to supply C 8 -ACP for lipoic acid biosynthesis (39)(40)(41) coupled with the low levels of these fatty acids in the PG molecular species grown in defined medium. C 7 -ACP is the first odd intermediate and arises from the FabH condensation of a C 5 -CoA with malonyl-ACP. C 15 -ACP was the next most abundant acyl-ACP, and it is the primary PlsC substrate. The high relative abundance of acyl-ACP intermediates suggests that FabF is the slow step in each round of elongation in S. aureus FASII. Figure 6. Alteration in acyl-CoA and fatty acid composition in media lacking branched-chain amino acids. Strain AH1263 was grown in defined media, and samples were taken at midlog phase. A, quantification of the CoA pool composition by LC-MS/MS using [ 13 C]C 2 -CoA as an internal standard. B, the weight percent of each fatty acid produced was determined by gas LC of the derived methyl esters quantified using a flame ionization detector. Triplicate biological replicates were analyzed. A two-tailed Student's t test was used to compare the two indicated groups. *p < 0.005.

Ile control of the acyl-ACP pool
The major shift in membrane fatty acid composition arising from the absence of Ile from the growth medium is the sharp increase in straight-chain even-carbon fatty acids. We compared the acyl-ACP pool composition of S. aureus grown in defined medium to growth in media lacking Ile and Leu. Even-carbon straight-chain acyl-ACP intermediates rose significantly in samples obtained from cells grown without Ile and Leu. This increase was coupled with a reduction in odd-chain acyl-ACP (Fig. 7B). These changes in pathway intermediate abundance reflect the alteration in fatty acid composition under these growth conditions. S. aureus FabH utilizes C 2 -CoA as a substrate (27,28,42), so FabH initiation with C 2 -CoA is the most likely origin of the elevated C 4 -ACP pool found in the absence of Ile and Leu. However, we cannot rule out an alternate pathway via acetyl-ACP condensation with malonyl-ACP by FabF, although how acetyl-ACP would be formed in S. aureus is unknown. The acyl-CoA:ACP transacylase activity of FabH may be responsible for acetyl-ACP, but there is no evidence for C 5 -ACP in the profile. C 5 -CoA is the preferred FabH substrate; therefore, C 5 -ACP should be present if there was significant FabH transacylation in vivo.

Discussion
Our results reveal the key role of branched-chain amino acid metabolism in determining the structure of membrane phospholipids in S. aureus (Fig. 8). Extracellular Ile and Leu have a major impact on the supply of FabH precursors to initiate fatty acid synthesis. The C 5 -CoA pool arises from the transamination of Ile and Leu followed by the formation of aC 5 -CoA and iC 5 -CoA by Bkd (Fig. 8). The ratio of aC 5 -CoA:iC 5 -CoA is 3.5 at equal extracellular Ile and Leu concentrations illustrating that Ile is the preferred substrate for the Bkd pathway. The Bkd selectivity for Ile is not stringent because removal of Ile from the media results in a large increase in iC 5 -CoA and iso-odd-chain fatty acids in membrane phospholipids. In contrast, Val is a poor substrate for the Bkd pathway and makes a minor contribution to fatty acid structure, even when it is the only extracellular branched-chain amino acid available (Fig. 8). The aC 5 -CoA:iC 5 -CoA ratio is 3.4 in the absence of extracellular Ile/Leu, meaning that the Bkd pathway has the same selectivity for substrates generated by de novo biosynthesis. C 2 -CoA is a major component of the acyl-CoA pool and arises from pyruvate dehydrogenase. Activation of acetate by C 2 -CoA synthetase and acetate kinase also supports the C 2 -CoA pool. Our metabolomics measurements show that the absence of branched-chain amino acids in the medium leads to a 75% decrease in the abundance of C 5 -CoA coupled with an increase in the amount of C 2 -CoA. This increases the C 2 :C 5 -CoA ratio from 0.86 in the presence of branched-chain amino acids to 5.7 in their absence leading to enhanced initiation of FASII using the C 2 -CoA primer. However, the selectivity of S. aureus FabH for C 5 -CoA (27,28,42) ensures there is always branched-chain acyl-ACP in the metabolic pipeline for the synthesis of either a15:0-ACP or i15:0-ACP for PlsC to place into the 2-position of the glycerol backbone. Although i14:0 derived from Val is used for the 2-position when Ile and Leu are absent from the medium, 15:0 carbon branched acyl chains dominate the 2-position of the phospholipid molecular species under all growth conditions studied. Our metabolomic datasets are consistent with the substrate selectivity exhibited by S. aureus FabH for C 2 -CoA and C 5 -CoA in vitro (27,28,42). In summary, our results point to the CoA thioester primer pool composition as being the primary determinant of the fatty acid composition of S. aureus. This means that changes in membrane phospholipid structure in S. aureus triggered by environmental conditions or signaling pathways are driven by changes in branched-chain amino acid metabolism that in turn impacts the acyl-CoA FabH primer pool.
Branched-chain amino acids are at the intersection of metabolism and virulence in S. aureus (43)(44)(45)(46)(47). Their levels (primarily Ile) are sensed by the CodY transcriptional regulator that controls a suite of genes involved in amino acid biosynthesis and virulence. When environmental Ile is limiting, the direct repression of alpha-toxin, hyaluronidase, and Panton-Valentine leucocidin by CodY is relaxed (48), and ΔcodY strains are hypervirulent (49). S. aureus encounters limited quantities of branched-chain amino acids at the infection site, where they compete with the host for these essential nutrients (43). Consistent with this view, S. aureus phospholipid Figure 8. Pathways responsible for the generation of short-chain acyl-CoA precursors. Pyruvate (Pyr) arising from glycolysis is converted to C 2 -CoA by pyruvate dehydrogenase (Pdh) and is the precursor for evennumbered straight-chain fatty acids. There are other sources of C 2 -CoA including C 2 -CoA synthetase and phosphotransacetylase that are not shown. Extracellular branched-chain amino acids are transaminated by IlvE giving rise to the respective ketoacids. Alternately, the ketoacids are generated by the de novo branched-chain amino acid biosynthetic pathway. These ketoacids are decarboxylated to their respective acyl-CoAs by branched-chain ketoacid dehydrogenase (Bkd) to form 2-methyl-butyryl-CoA (aC 5 -CoA) and isobutyryl-CoA (iC 5 -CoA). Valine is a poor substrate for the Bkd pathway, whereas Ile is the major amino acid metabolized. The ratio of short-chain acyl-CoA primers available to FabH is a major determinant of fatty acid composition. The anteiso branched-chain odd-carbon fatty acids arise from aC 5 -CoA, iso branched-chain odd-carbon fatty acids are from iC 5 -CoA, straight-chain even-carbon fatty acids come from C 2 -CoA, and the trace of iso branched-chain even-carbon fatty acids are derived from Val. structure at the infection site is mimicked in the laboratory by a low Ile environment (36,37). Similar effects on fatty acid composition are noted in S. aureus strains lacking branchedchain amino acid transporters (15) that reduce the availability of Ile for FASII and impacts virulence (50). Exogenous straight-chain fatty acids are incorporated into the 1-position of S. aureus phospholipids and suppress straight-chain synthesis (24,33). This effect also arises when Ile is limiting and likely driven by changes in the supply of FabH primers. In future work studying the adaption of membrane composition to changes in the environment, measurement of the FabH precursor pool composition will be a requisite experiment to determine the impact of altered branched-chain amino acid metabolism. Whereas the ratio of straight:anteiso:iso fatty acids is governed by the composition of the acyl-CoA primer pool, chain length control remains a property governed by FASII components and their interaction with acyltransferases ( Fig. 1).  15 N has no impact on the experiment. [ 13 C 2 ] C 2 -CoA was from Millipore Sigma. Defined media components and other reagents were purchased from Millipore Sigma or Thermo Fisher Scientific. All MS reagents are of HPLC grade or better. The defined media used in this work consist of M9 salts supplemented with amino acids and other nutrients as outlined in Table S2. The growth characteristics of S. aureus strain AH1263 in defined media lacking each of the branched-chain amino acids are illustrated in Figure S4. Like other S. aureus strains, wildtype strain AH1263 requires extracellular Val as a supplement in defined media (47,51). Growth when Leu was removed was slower than in the other conditions as reported previously, but removal of either Ile or Ile + Leu did not affect growth (47,51) (Fig. S4). The Leu growth phenotype arises from the Ile repression of branchedchain amino acid synthesis by CodY (47).

Branched-chain amino acid growth experiments
An overnight culture of strain AH1263 was grown in defined medium at 37 C (Table S2). The cells were collected by centrifugation at 4000g for 10 min at room temperature, washed with defined medium lacking Ile, Leu, and Val, and again collected by centrifugation. The cell pellets were resuspended in defined medium lacking Ile, Leu, and Val. The cell suspensions were used to inoculate strain AH1263 into the defined medium being used for the experiment to an absorbance of 0.05 at 600 nm. The growth rates of strain AH1263 in the presence or the absence of branched-chain amino acids are shown in Figure S4. In the metabolomics experiments, the cell cultures were grown to early log phase at an absorbance of 0.6 at 600 nm, at which point the cells were collected and extracted using the various workflows for the analysis of the different metabolomes.
Cells were labeled with heavy amino acids by collecting cells grown at 37 C to an absorbance of 0.5 at 600 nm in defined medium by centrifugation at 4000g for 10 min at room temperature. The cell pellets were washed with defined medium lacking Ile, Leu, and Val and were collected by centrifugation. The cell pellets were resuspended in defined medium with the normal concentrations of Ile, Leu, and Val being replaced by [ 13 C 6 ]isoleucine, [d 3 ]leucine, and [ 13 C 5 , 15 N]valine. In the second labeling experiment, the cells were split in medium containing 50 μg/ml [d 3 ]leucine and 20 μg/ml [ 13 C 5 , 15 N]valine either with or without 50 μg/ml Ile. These cultures were incubated at 37 C, and samples were taken at 10, 20, or 40 min for metabolomic analyses.

Acyl-CoA MS
The cell pellet from 10 ml of culture was resuspended in 2 ml methanol and 1 ml water and incubated on ice for 30 min. Chloroform (1.5 ml) and water (1.2 ml) were added to remove the lipids. The aqueous top layer was collected, and 30 pmol of [ 13 C 2 ]C 2 -CoA (Millipore Sigma) was added, and the samples were dried overnight in a Savant Speedvac Concentrator SPD 1010 (Thermo Fisher Scientific). The dried sample was resuspended in 300 μl water, and 20 μl aliquots were used for analysis. Acyl-CoAs were analyzed using a Shimadzu Prominence UFLC attached to a QTrap 4500 equipped with a Turbo V ion source (Sciex). Samples were injected onto an Acquity UPLC HSS C18, 2.5 μm, 3.0 × 150 mm column at 40 C (Waters) using a flow rate of 0.2 ml/min. Solvent A was 10 mM ammonium acetate, pH 6.8, and solvent B was 95% acetonitrile + 10 mM ammonium acetate, pH 6.8. The HPLC program was the following: starting solvent mixture of 83% A/17% B, 0 to 2 min isocratic with 17% B; 2 to 15 min linear gradient to 50% B; 15 to 20 min linear gradient to 95% B; 20 to 29 min isocratic with 95% B; 29 to 30 min linear gradient to 17% B; and 30 to 35 min isocratic with 17% B. The QTrap 4500 was operated in the positive mode, and the ion source parameters were ion spray voltage, 5500 V; curtain gas, 15 psi; temperature, 400 C; collision gas, medium; ion source gas 1, 15 psi; ion source gas 2, 20 psi; declustering potential, 60 V; and collision energy, 45 V. Multiple reaction monitoring (MRM) is a specific and sensitive technique using a triple quadrupole mass spectrometer to quantify compounds in complex mixtures. MRMs are reported as two masses. The Q1 mass is for the intact molecule, and the Q3 mass is for the daughter ion being monitored.

Lipid MS
Lipids were extracted from 5 ml of culture (52), and the dried lipids were suspended in chloroform/methanol (1:1). PG was analyzed using a Shimadzu Prominence UFLC attached to a QTrap 4500 equipped with a Turbo V ion source. Samples were injected onto an Acquity UPLC BEH HILIC, 1.7 μm, 2.1 × 150 mm column (Waters) at 45 C with a flow rate of 0.2 ml/min. Solvent A was acetonitrile, and solvent B was 15 mM ammonium formate, pH 3.0. The HPLC program was the following: starting solvent mixture of 96% A/4% B; 0 to 2 min, isocratic with 4% B; 2 to 20 min, linear gradient to 80% B; 20 to 23 min, isocratic with 80% B; 23 to 25 min, linear gradient to 4% B; and 25 to 30 min, isocratic with 4% B. The QTrap 4500 was operated in the Q1 negative mode. The ion source parameters for Q1 were as follows: ion spray voltage, −4500 V; curtain gas, 25 psi; temperature, 350 C; ion source gas 1, 40 psi; ion source gas 2, 60 psi; and declustering potential, −40 V. The system was controlled by the Analyst software. The sum of the areas under each peak in the mass spectra was calculated, and the percentage of each molecular species present was calculated with LipidView software (Sciex).
Samples were also introduced to the QTrap 4500 by direct injection to perform product scans to verify the fatty acids present in a particular PG molecular species. The ion source parameters for negative mode product scan were ion spray voltage, −4500 V; curtain gas, 10 psi; collision gas, medium; temperature, 270 C; ion source gas 1, 10 psi; ion source gas 2, 15 psi; declustering potential, −40 V; and collision energy, −50 V.

Acyl-ACP MS
Samples were prepared for acyl-ACP quantification as previously described (53) with modifications. One milliliter aliquots of cultures at midlog phase growth were removed to a 1.5 ml Eppendorf tube containing 250 μl ice cold 10% trichloroacetic acid. The tubes were inverted and kept on ice. Harvested cells were washed with acetone, dried, and then stored at −80 C. Pellets were resuspended in 100 μl of lysis buffer (50 mM sodium phosphate buffer, pH 7.2, 1 mM ascorbic acid, 2 mM EDTA, and 6 M urea). Lysis buffer was prepared just prior to use. [ 13 C 2 ]Acetyl-ACP was added as an internal standard (1 μl of 0.0125 μg/μl [ 13 C 2 ]acetyl-ACP). A chloroform/methanol extraction was performed by adding 400 μl methanol, followed by 100 μl of chloroform. Samples were sonicated in a sonication bath (Fisher Scientific CPXH Series 1.9L) for 10 min at room temperature to resuspend. Phases were separated by adding 300 μl of 0.2 mM formate buffer, pH 3.9. The upper phase was discarded, and 300 μl of methanol was added to the remaining sample. The precipitated proteins were collected by centrifugation for 5 min at top speed in an Eppendorf microfuge (20,000g), and the pellets were washed with 300 μl of methanol and dried. The dried pellets were resuspended in 10 μl of 100 mM sodium phosphate buffer, pH 6.5, and sonicated for 10 min at room temperature in an ultrasonic bath. Insoluble debris was removed by centrifugation at top speed in a microfuge. A 5 μl aliquot of the supernatant was removed to a clean tube. Proteins were digested by adding 0.5 μg endoproteinase Asp-N (Millipore Sigma) in 10 μl of 100 mM Tris, pH 7.5 (38). Samples were incubated 1 h at room temperature. Proteolysis was quenched with 15 μl methanol.
Acyl-ACP species were analyzed using a Shimadzu Prominence UFLC attached to a QTrap 4500 equipped with a Turbo V ion source. Samples were injected onto an Acquity UPLC Discovery B10 Wide Pore C18, 3 μm, 2.1 × 100 mm column (Millipore Sigma) at 25 C with a flow rate of 0.2 ml/min. Solvent A was 10 mM ammonium formate, pH 3.6, in 10% acetonitrile, and solvent B was 10 mM ammonium formate, pH 3.6, in 90% acetonitrile. The HPLC program was the following: starting solvent mixture of 100% A/0% B; 0 to 4 min, linear gradient to 10% B; 4 to 12 min, linear gradient to 100% B; 12 to 17 min, isocratic with 100% B; 17 to 20 min, linear gradient to 0% B; and 20 to 25 min, isocratic with 0% B. The QTrap 4500 was operated in the positive mode, and the ion source parameters were ion spray voltage, 5500 V; curtain gas, 30 psi; temperature, 400 C; collision gas, medium; ion source gas 1, 25 psi; and ion source gas 2, 30 psi. The Q1/Q3 MRM parameters for [ 13 C 2 ]acetyl-ACP were 718.3/350.2 with a declustering potential of 80 V and a collision energy of 40 V. The analytic parameters for all the ACP pathway intermediates are listed in Tables S3-S6. The system was controlled by the Analyst software and analyzed with MultiQuant 3.0.2 software (Sciex). [ 13 C 2 ]C 2 -ACP was used as the internal standard. Peaks corresponding to individual acyl-ACP species were quantified relative to the internal standard.

Fatty acid gas chromatography
Fatty acid methyl esters were prepared using methanol/hydrochloric acid. The fatty acid methyl esters were analyzed by a Hewlett-Packard model 5890 gas chromatograph equipped with a flame ionization detector and separated on 30 m × 0.536 mm × 0.50 μm DB-225 capillary column (Agilent). The injector was set at 250 C, and the detector was at 300 C. The temperature program was as follows: initial temperature of 70 C for 2 min, rate of 20 C/min for 5 min (final 170 C), rate of 2 C/min for 10 min (final 190 C), hold at 190 C for 5 min, rate of 2 C/min for 15 min (final 220 C), and hold at 220 C for 5 min. The identity of fatty acid methyl esters was determined by comparing their retention times with fatty acid methyl ester standards (Matreya). The compositions were expressed as weight percentages.
Preparation of [ 13 C 2 ]acetyl-ACP 13 C-labeled acetyl-ACP was synthesized by the previously described method for crotonyl-ACP synthesis (54) with slight modifications. His 6 -tagged E. coli apo-ACP was expressed using the pET-15b plasmid transformed into BL21(DE3) cells. After induction with 1 mM IPTG for 3 h at 37 C, cells were lysed and apo-ACP was purified using nickel affinity chromatography (GoldBio). Pooled purified protein was dialyzed overnight in 20 mM Tris, pH 7.5, to remove the imidazole. The His 6 tag was then cleaved by thrombin treatment EDITORS' PICK: Control of S. aureus fatty acid composition (Millipore) and separated from the apo-ACP by nickel chromatography. [ 13 C 2 ]Acetyl-ACP was synthesized by incubating 300 μM purified apo-ACP with 2 μM E. coli ACP synthase and 600 μM [ 13 C 2 ]acetyl CoA (Millipore Sigma) in 50 mM Tris, pH 7, with 10 mM MgCl 2 for 2 h at 37 C. The His 6 -tagged [ACP]synthase was subsequently removed from the reaction by nickel affinity chromatography. The MgCl 2 and excess CoA were removed using a PD-10 desalting column. The [ 13 C 2 ] acetyl-ACP was eluted in 20 mM Bis-Tris, pH 6, with 200 mM NaCl and concentrated using a 3 kDa cutoff Amicon Ultra centrifugal filter. The concentrated standard was quantified using the Bicinchoninic Acid Protein Assay Kit (Pierce).

Statistical analysis
Measurements are means ± SEM of biological replicates. The Student's t test was used to compare two groups using GraphPad Prism software (version 9.1.2; GraphPad Software, Inc).

Data availability
All study data are included in the article.
Supporting information-This article contains supporting information.
Acknowledgments-We thank Laura Creed and Pam Jackson for their expert technical assistance. Funding and additional information-This work was supported by the National Institutes of Health grant GM034496 (to C. O. R.), Cancer Center Support grant (CA21765), and the American Lebanese Syrian Associated Charities. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Conflict of interest-The authors declare they have no conflicts of interest with the contents of this article.