Results
Mutations in the mbcS promoter restore growth of alpdA mutant during laboratory growth. A lpdA mutant lacks the dihydrolipoamide dehydrogenase subunit of the BKDH complex and is deficient for BCFA synthesis (Teoh et al., 2021). As expected, the strain is not viable in rich, complex medium (tryptic soy broth [TSB]) and does not achieve an optical density beyond ~0.1 after ~7-8 h unless supplemented with a mixture of the branched-chain carboxylic acids (BCCAs; 2-methylbutyric acid [a C5], 3-methylbutyric acid [i C5], and isobutyric acid [i C4]) or with a 15:0 BCFA (Fig 2A-C , compare squares to circles) . However, we observed that the lpdA mutant formed small colonies upon prolonged incubation on tryptic soy agar (TSA) in the absence of BCCAs. We reasoned that these colonies harbor additional mutations that bypass the requirement for the BKDH complex. To test this, independently isolated mutants were submitted for whole genome sequencing and the sequences of the apparent suppressor mutants and the parent strain were compared with BRESEQ (Barrick et al., 2009). Interestingly, nearly all mutants harbored either single nucleotide polymorphisms or insertion-deletions in the promoter region of SAUSA300_2542,which is annotated to encode an acyl-CoA synthetase (Table 1, Fig 3A). We refer to these alleles hereafter as mbcS alleles to reflect the m ethylb utyric acid supplementation bypassed in these strains. We first analyzed growth behavior of a representative lpdA mutant with a mutant mcbS allele (i.e., mbcS1 ). This strain achieved a doubling time and cell yield essentially identical to that of the WT strain, even in the absence of BCFAs or their carboxylic acid precursors (Fig 2A-C,compare triangles and squares to circles) .
A recent study reported that TSB and other standard laboratory media contain trace amounts of branched-chain carboxylic acids (Whaley et al., 2023) . We wondered whether mbcS was required to scavenge these BCFA precursors and promote BCFA synthesis in the absence of lpdA . To test this, we constructed an lpdA mbcS double mutant using available mutants from the Nebraska Transposon Mutant Library (Fey et al., 2013) and compared the strain’s growth behavior to the lpdA single mutant. Unlike the lpdA mutant, whose growth was restored with branched-chain carboxylic acids, the lpdA mbcS double mutant failed to grow beyond an optical density of ~0.1 unless a 15:0 was included in the medium (Fig 2 , compare inverted triangles to squares). These data indicate that the mutations in the mbcS promoter suppress the BCFA auxotrophy, and that mbcS is required to restore growth when BCCAs are provided exogenously in the growth medium.
Promoter-up mutations inmbcS alter the membrane fatty acid profile in S. aureus.Many of the changes we observed in the mbcS promoter of our mutants either decrease the spacing between the -35 and -10 boxes from 18 nucleotides to the typical 17 nucleotides found in S. aureus , or refine the -35 and -10 sequences toward a consensus σA-dependent promoter (Table 1, Fig 3A)(Helmann, 1995). We hypothesized that these changes would increasembcS expression. To test this, we introduced a gfpreporter plasmid containing either the wild-type (WT) promoter ofmbcS or the promoters isolated from two independently isolated mutants into wild-type cells and measured promoter activity (Fig 3A, B) . During growth in TSB, mbcS expression is relatively low in WT cells. Mutation of the mbcS promoter resulted in a 3- to 4-fold increase in promoter activity (Figure 3B) . We then wondered whether overexpressing mbcS altered the membrane fatty acid profile, supporting growth in TSB. To address this question, we grew the WT and our two selected lpdA suppressor mutants to exponential phase and subjected the cells to gas chromatography fatty acid methyl ester analysis (GC-FAME). As expected, WT cells have a relatively high proportion of a 15:0 BCFAs in their membranes with lower levels of i 14:0 BCFAs. Interestingly, we measured a large increase in i 14:0 BCFAs and relatively low levels of a 15:0 BCFAs in the suppressor mutants (Fig 3C ). These data strongly suggest that overexpressing mbcS rescues BCFA auxotrophy by promoting the synthesis of iso but not anteiso BCFAs.
MbcS is a branched-chain acyl-CoA synthetase. SAUSA300_2542 is annotated to encode an acyl-CoA synthetase we call MbcS. Three facts support this annotation. First, MbcS is 56% similar to (end-to-end)S. aureus acetyl-CoA synthetase (Burckhardt et al., 2019). Second, MbcS is 43-47% similar to other bona fide acyl-CoA synthetases from Salmonella enterica and Rhodopseudomonas palustris(Crosby & Escalante-Semerena, 2014; Crosby et al., 2012; Starai & Escalante-Semerena, 2004). Third, close inspection of the predicted amino acid sequence of MbcS reveals a conserved motif and catalytic lysine residue among representative members of the AMP-forming family of acyl-CoA synthetases (Fig 4A ) (Burckhardt et al., 2019; Crosby & Escalante-Semerena, 2014; Crosby et al., 2012; Starai et al., 2002). To begin to test the hypothesis that MbcS has acyl-CoA synthetase activity, we complemented a lpdA mbcS double mutant with integrative plasmids coding for either the wild-type S. aureus mbcS (SambcS +) or a variant substituting the catalytic lysine residue with alanine (K510A) under the control of an anhydrotetracycline inducible promoter and the Tet repressor (TetR) and assessed growth behavior in TSB. When we introduced the empty vector into the mbcS lpdA double mutant, the cells were unable to achieve an optical density beyond ~0.1 after ~6-8 h (Fig 4B , compare triangles to circles) . Growth was restored in this mutant when we introduced a wild-type copy of mbcS but not the mutant allele coding for the catalytically inert MbcSK510A variant, strongly suggesting that MbcS-dependent acyl-CoA synthetase activity was necessary and sufficient for growth in TSB (Fig 4B , compare inverted triangles and diamonds to circles) . Interestingly, we measured no obvious growth defect in the single mbcS mutant. To test genetically if MbcS indeed supports growth of the lpdAmutant by catalyzing acyl-CoA synthetase activity, we complemented thelpdA mbcS double mutant with an integrative plasmid coding for the bona fide acyl-CoA synthetase IbuA from R. palustris(RpibuA+ ), and assessed growth behavior(Fig 4C, D) . Unlike SambcS+ ,RpibuA+ does not complement the lpdA mbcS mutant for growth in TSB in the absence of inducer (Fig 4C) . However, adding 25 ng ml-1 of anhydrotetracycline (aTc) to the growth medium was sufficient to restore growth of the lpdA mbcS strain to a level similar to the WT strain (Fig 4D) . Taken together, our genetic data indicate that, in the absence of the BKDH complex, the acyl-CoA synthetase activity of MbcS bypasses the BCFA auxotrophy of the lpdAmutant.
To support the in vivo data above, we cloned, overexpressed, and purified N- terminal hexahistidine (His6)-tagged MbcS protein (Sa MbcS). Following tag removal, we reisolated the protein to apparent homogeneity and conducted a coupled spectrophotometric assay with Mg*ATP, coenzyme A (CoA) and a variety of carboxylic acids as substrates as described in Experimental ProceduresSa MbcS activated short, straight and branched-chain monocarboxylic acids and typically preferred branched substrates over straight substrates with the same carbon length. Sa MbcS had the highest activity with isobutyrate (i C4)(Fig 5A) . We saw no activity when a 15:0 was used as substrate in the reaction (data not shown).
The coupled assay does not provide a direct identification of the product formed. Therefore, we used LC-MS analysis to authenticate isobutyryl-CoA (i C4-CoA) formation from isobutryate (i C4). Briefly, we incubatedSa MbcS with Mg*ATP, isobutyrate (IB), and CoA in buffer. As a negative control, Sa MbcS was denatured at 85°C, and its activity was tested. Using LC-MS, we successfully separated, detected, and confirmed the identities of isobutyryl-CoA and CoA in reaction mixtures containing recombinant Sa MbcS. We failed to detect IB-CoA formation when substrates were incubated with denatured MbcS, demonstrating product formation was enzyme-dependent (Fig 5B) .
We then determined a limited set of kinetic parameters for Sa MbcS and compared them to bona fide isobutyryl-CoA synthetase from R. palustris (Rp IbuA) using the same continuous spectrophotometric assay. Because our interest lay in the specificity of the enzyme for substrates relevant to BCFA synthesis, we focused on carboxylic acid substrates and used a fixed amount of coenzyme A and ATP co-substrates. Under our in vitro conditions, the apparent affinity ofSa MbcS for i C4 anda C5 were essentially identical (Km values were calculated to be 8.9 ± 2.1 μM and 5.5 ± 1.0 μM, respectively). Sa MbcS displayed a significantly lower affinity for i C5 (> 500 μM) (Table 2) . MbcS displayed a measurably higher affinity for i C4 compared toRp IbuA. This seemed to be offset by the faster rate of reaction catalyzed by Rp IbuA compared to Sa MbcS usingi C4 as substrate (Table 2 , compare Kcat values for Rp IbuA andSa MbcS) . This resulted in comparable catalytic efficiencies for the two enzymes (compare Kcat/Km values). As expected, the catalytically inert Sa MbcSK510A variant exhibited no detectable acyl-CoA synthetase activity (Table 2) . Taken together, these data indicate that Sa MbcS is a bona fide methylbutyryl-CoA synthetase with a preference fori C4 as substrate and participates in branched-chain fatty acid synthesis.
mbcS supports the utilization of 2-methylbutyraldehyde as a precursor for BCFA synthesis. The BKDH complex catalyzes an oxidative decarboxylation reaction, converting the branched-chain α-keto acids to their cognate BCCAs. These BCCAs are activated to acyl-CoAs that feed fatty acid synthesis (Frank et al., 2021; Sen et al., 2015; Singh et al., 2008; Ward et al., 1999). Blocking BKDH function revealed a second enzyme that performs this activation step. S. aureus produces aldehydes during isoleucine, leucine, and valine catabolism in peptide-rich environments (Bos et al., 2013; Filipiak et al., 2012). These aldehydes are conceivably oxidized to carboxylic acids and activated by MbcS. Indeed, staphylococci used in the food industry, such as S. carnosus , oxidize branched-chain aldehydes to their respective BCCAs to synthesize BCFAs (Beck, 2005; Beck et al., 2002). To test if this is also true for S. aureus , we grew cells in chemically defined medium (CDM) (Sheldon et al., 2014) supplemented with the isoleucine-derived aldehyde 2-methylbutyraldehyde (2MA), the BCCA 2-methylbutyric acid (a C5), and the BCFAa 17:0. As expected, the lpdA single mutant failed to achieve a final optical density beyond ~0.1 in the absence of supplementation. Both the lpdA mutant and thelpdA mbcS1 suppressor mutant grew to similar levels in medium containing the aldehyde and the carboxylic acid (albeit not to WT levels for the lpdA mbcS1 strain). BCCA- and aldehyde-dependent growth was dependent on mbcS (Fig 6 , compare pink and orange bars, lpdA::kan+ mbcS:: φNΣ vs.lpdA::kan+ ) . Indeed, the double mutant only grew in the presence of a 17:0 BCFA, indicating that lack of viability was due to insufficient intracellular branched-chain acyl-CoA precursors.
S. aureus heavily suppresses branched-chain amino acid (BCAA) synthesis during rapid growth (Kaiser et al., 2018; Waters et al., 2016). Rather, BCAAs are preferentially imported during growth as either free amino acids or as peptides and play critical roles for cell physiology. As abundant amino acids found in proteins, BCAAs can be used directly for protein synthesis and drive folding via the hydrophobic effect (Brosnan & Brosnan, 2006; Dill 1990). BCAAs are also readily interconverted to their α-keto acids that serve as precursors for pantothenate and coenzyme A, and for the BCFAs. Thus, the metabolism of BCAAs controls membrane phospholipid composition (Frank et al., 2021; Richardson et al., 2015). Lactococcus lactis strains are capable of decarboxylating BCAAs, particularly leucine, to their corresponding aldehyde 3-methylbutanal. These aldehydes are important for flavor formation in certain cheeses by Lactococcus spp. (Rijnen et al., 2003; Smit et al., 2005; Smit et al., 2004). We wondered whetherS. aureus might also be capable of generating BCFAs de novo through a pathway parallel to that formed by the BKDH complex. To test this, we compared the growth behavior of the lpdA mutant to the WT parent strain, with and without mbcS overexpressed. Unlike the lpdA mutant, the lpdA mbcS1 mutant grew, albeit poorly, in CDM medium without BCFA or precursor supplementation. Again, this growth was abrogated in the lpdA mbcS double mutant(Fig 6 , grey and black bars, mutant strains compared to WT) .
Metabolites from S. epidermidis restore growth oflpdA mutant in an MbcS-dependent manner. The human skin is an important habitat for microbial colonization. Metagenomic studies show that different species of bacteria, fungi and viruses compose the skin microbiota (Byrd et al., 2018). Coagulase negative staphylococci (CoNS) such as S. epidermidis , S. hominis , S. haemolyticus , and S. lugdunensis are the most abundant skin colonizers and serve as a barrier to protect against local and systemic infections caused by pathogenic species like S. aureus (Brown & Horswill, 2020; Byrd et al., 2018; Zipperer et al., 2016). Indeed, skin commensals successfully compete with S. aureus when resources become limited in the skin environment. For instance, S. lugdunensis synthesizes the antimicrobial peptide lugdunin that effectively inhibits the growth of S. aureus in vitro and interferes with nasal colonization in a mouse model (Zipperer et al., 2016).
Microbial cells within natural communities also can exchange metabolites with one another, working together to distribute labor before interactions potentially become antagonistic (Pande et al., 2014). Considering the interactions between S. aureus and staphylococci in the skin, we wondered if S. epidermidis could feed ourS. aureus lpdA mutant given its propensity to make volatile compounds derived from the branched-chain amino acids. To address this, we streaked the lpdA single mutant on medium lacking branched-chain fatty acids next to S. epidermidis. IfS. epidermidis can cross-feed S. aureus, we expect to observe growth adjacent to S. epidermidis, with the zone supporting growth proportional to the amount of the nutrient being secreted and diffusing through the medium. Interestingly, thelpdA mutant grew only in close proximity to S. epidermidis, and this growth was abrogated when we streaked thelpdA mbcS double mutant next to S. epidermidis(Fig 7A ). We then quantified the effect using S. epidermidis conditioned medium. As expected, the lpdA single mutant was chemically complemented with a mix of branched-chain carboxylic acids or a 17:0 BCFAs (Fig 7B , compare grey, orange bars to black bars, respectively) . We then determined that 10% conditioned medium from S. epidermidis restored near WT growth. The effect was lost when the cells were supplemented with decreasing concentrations of conditioned medium (Fig 7B , compare green, pink, and blue bars). Only a 17:0 supplementation restored growth of the lpdA mbcS double mutant(Fig 7B) . Taken together, these data indicate S. aureuscan salvage isoleucine-derived, exogenous branched-chain aldehydes in addition to branched-chain carboxylic acids in a mbcS- dependent manner for BCFA synthesis. Our data also strongly suggest that S. epidermidis can support S. aureus growth by providing these same essential precursors for BCFA synthesis.