Construction and confirmation of a ΔplsX in-frame mutant

plsX is the second gene in a multicistronic operon also comprising FASII genes (S1 Fig). We constructed in-frame ΔplsX mutants of RN-R (RN4220 repaired for a defective fakB1 gene in the 8325 lineage [15]) and of USA300_FPR3757 JE2 strains. Both ΔplsX mutants failed to grow on BHI solid medium in the absence of FAs. Deletions were confirmed by PCR and whole genome sequencing (S1 Table) and by RN-R ΔplsX growth complementation in the absence of FAs using a plasmid-carried copy of the intact plsX gene (S2 Fig).

FA auxotrophy of S. aureus ΔplsX mutants is suppressed by a point mutation in fabF

During phenotypic characterization of the JE2 ΔplsX mutant, we noted the appearance of colonies on BHI medium without added FA (Fig 2A, left plate). Note that BHI reportedly contains traces of FA [16], but which are insufficient to confer ∆plsX growth. Two colonies subjected to whole genome sequencing confirmed the in-frame plsX deletion and also revealed point mutations mapping to FabF (SAUSA300_0886), the 3-oxoacyl-ACP synthase II of the FASII elongation cycle (see Fig 1), producing FabF^A119E (alanine to glutamic acid; Sup1) and FabF^D266A (aspartic acid to alanine; Sup2). The suppressor carrying FabF^D266A was not recoverable from the primary stocked strain, possibly because the mutation was deleterious for growth. Sup1 and Sup2 also carried the same point mutation in csa1A (SAUSA300_0100), encoding a tandem lipoprotein.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. fabF mutants or the FabF inhibitor platensimycin rescues ΔplsX growth in the absence of exogenous FAs.

A. Lawns of RN-R ΔplsX were spread on BHI medium without FA addition, without (left) or with (right) 1.5 µg platensimycin (P) deposited in the center. Plates were photographed after 4 days at 37°C. Black arrowheads, examples of suppressors, some of which map to fabF. Dashed arrow (right), a homogenous growth ring surrounds the platensimycin spot. N = 4. B. Working model for ΔplsX growth rescue by a FabF mutation (FabFmut) or by partial FabF inhibition by platensimycin (Anti-FabF). FabF perturbations would reduce enzyme processivity, such that the unstable FabF-FA intermediate releases FAs prior to second substrate malonyl-ACP loading, as needed to bypass the ΔplsX defect. FabF in blue; FAs in green. Symbols for FabF are based on [51].

https://doi.org/10.1371/journal.pgen.1012165.g002

We expanded our analyses by isolating 31 other ΔplsX suppressors from RN-R or JE2 backgrounds (16 and 15 suppressors, respectively), and screened for those carrying fabF mutations by PCR followed by DNA sequence analyses (S2 Table). Among them, 14 carried fabF mutations, which mapped to the same fabF nucleotide position in both strain backgrounds, resulting in a FabF^A119E alteration as in Sup1. This allele is referred to as fabF1. csa1A, which carried a mutation in Sup1 and Sup2 mutants, was not mutated in the 14 fabF mutants, indicating that fabF was the suppressor allele. Emergence of the same fabF mutation in both suppressor strain backgrounds could suggest that few FabF point modifications would allow both FASII activity and ∆plsX rescue.

We attempted to prove that the fabF mutation was directly responsible for the ΔplsX suppressor phenotype by PCR-amplifying the WT fabF and mutant fabF1 alleles, cloning on a medium copy-number vector and constitutive promoter, and establishing plasmids in the ΔplsX and/or fabF1ΔplsX strains, with the goal of reversing the suppressor phenotype. A clone expressing fabF1, but not WT fabF could be established in the WT RN-R strain by direct transformation.; this is consistent with the putative reduced activity of this allele. We further failed to establish the fabF1 clone in the ΔplsX background, even in FA-supplemented medium. This suggested that fabF overexpression was toxic in S. aureus, as reported in Escherichia coli [17], and that our strategy was not feasible.

As an alternate approach, we used a FabF inhibitor, platensimycin, to determine whether disabling FabF was responsible for ΔplsX suppression. This antibiotic binds to the acyl-ACP-FabF intermediate, and competes with malonyl-ACP entry, thus preventing FA elongation [18]. However, it also leads to increased levels of FabF and other FASII enzymes in Bacillus subtilis [19]. We tested whether platensimycin, like the FabF mutant, would overcome the ΔplsX growth defect, presumably by slowing FabF activity. RN-R ΔplsX cultures were plated on BHI solid medium, and platensimycin (1.5 µg) was deposited on plates, which were then incubated 96 h at 37°C (Fig 2A, right plate). In addition to the emergence of ΔplsX suppressors, a distinct homogeneous growth ring appeared around the platensimycin spots. Similar experiments were performed in liquid medium using 2-fold dilutions of platensimycin concentrations ranging from 1 ng to 500 ng/ml (S3 Fig). Compared to BHI medium, growth between 7 and 12 h was significantly stimulated by the presence of 125 ng/ml platensimycin (P=≤0.05). However, experiments in liquid medium are potentially flawed as mutant emergence and platensimycin stimulation cannot be distinguished. Ring formation on solid medium appears to be a more accurate means of assessing growth stimulation. Since the fabF mutants and a FabF inhibitor both restore ΔplsX growth, we consider it likely that the FabF mutation is directly responsible for ΔplsX suppression. Reducing fabF efficacy by mutation or subinhibitory platensimycin treatments could destabilize the FabF-acyl-ACP intermediate, to liberate free FAs and complement the ΔplsX growth defect (Fig 2B).

FA auxotrophy of S. aureus ΔplsX mutants is suppressed by mutations in fadM that map to the acyl-CoA binding cavity

We searched for mutations that conferred ΔplsX growth on BHI among the 17 remaining suppressors carrying a wild type fabF sequence. Genomic DNA sequencing was performed on three such isolates (one derived from RN-R ΔplsX and two from JE2 ΔplsX backgrounds). A single common gene target, SAOUHSC_01348 in RN-R or SAUSA300_1247 in JE2, was mutated (S1 and S2 Tables). While annotated as a 1,4-dihydroxy-2-naphthoyl-coenzyme A (DHNA-CoA) thioesterase implicated in menaquinone synthesis, the enzyme was shown to have acyl-CoA thioesterase activity [20]. Moreover, the 155-amino-acid ORF shares ~40% similarity and a conserved serine, histidine and aspartic acid catalytic triad with FadM, an E. coli acyl-CoA thioesterase and ACP binding protein (PBD database protein 1NJK), [20–24]. The S. aureus gene, referred to as fadM, was PCR-amplified and sequenced in the 17 other suppressors lacking mutations in fabF. All carried mutations in fadM. A single variant (FadM^Y90F, [encoded by fadM2]) was identified in the four isolates selected from RN-R ΔplsX; three distinct FadM variants were identified among the JE2 ΔplsX suppressors (FadM^I38T [fadM1], FadM^Y90F [fadM2], and FadM^Y133F [fadM3]) (S1 and S2 Tables). These results point to FabF and FadM as the main suppressors rescuing ΔplsX growth.

The S. aureus FadM crystal structure (PDB: 6FDG) and Alphafold predictions (https://alphafold.ebi.ac.uk/entry/A0A0H2XFE3) provide detailed information for mapping the identified FadM variants conferring ∆plsX suppression. The variants coincided with or mapped adjacent to amino acids reportedly involved in forming an FA binding cavity that favors long FA substrates [20] (Fig 3A). The grouped location of these variants leads us to speculate that acyl-CoA thioesterase activity may be defective in the fadM suppressors.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. FadM variants mapping to the predicted FA binding tunnel, but not FadM inactivation, confer the fadM suppressor phenotype.

A. Schematized S. aureus FadM monomer binding to FA, based on crystal structure and Alphafold predictions ([20,38], designed with https://alphafoldserver.com and http://www.cgl.ucsf.edu/chimera [52]). FA moiety (black line) binding in the FadM cavity involves amino acids Ile38 (pink), Tyr45; Met48, Leu122, Tyr125, and Phe126 (purple) [20]. The ∆plsX suppressors mapping to FadM affect amino acids Ile38Thr (pink), Tyr90Phe, or Tyr133Phe (red), which cluster around the FA-binding cavity; Ile38 (pink) is common to structural predictions and the FadM^I38T ∆plsX suppressor. B. FadM inactivation abolishes ΔplsX suppression: Growth of ΔplsX fadM2 and the ΔplsX fadM::Tn derivative was compared on solid BHI medium without or with C18:1. Plates were photographed after 48 h growth at 37°C. Background growth on BHI (left) may be due to FA carryover from pre-cultures or to FA traces in medium. Plates are representative of biological triplicates. C. Working model for FadM versus FadM mutant function. Binding of the acyl-CoA FA moiety (zigzag line) in the FadM tunnel is needed for acyl-CoA thioesterase activity (scissor). FadM mutations in the tunnel (red dot) prevent this activity (red X) but conserve ACP binding (thick arrow, lower right).

https://doi.org/10.1371/journal.pgen.1012165.g003

If FadM acyl-CoA thioesterase activity were needed for ∆plsX suppression, then its substrate, acyl-CoA, would need to be produced. In S. aureus, acyl-CoA is synthesized by acyl-CoA synthetase (encoded by fadE), which is part of the recently elucidated FA degradation (fad) system (Fig 1; [25,26]). The fad genes are subject to CcpA-mediated glucose repression [25–27]. However, glucose additiondid not affect the fadM mutant ∆plsX suppressor phenotypes (S4 Fig). Based on above results, we consider it reasonable to rule out a role for FadM acyl-CoA thioesterase activity in ∆plsX growth rescue.

A fadM null mutant does not confer ΔplsX growth

Suppression by fadM point mutants could be due to FadM-mediated ACP binding, or to the loss of all FadM activity. We asked whether inactivation of fadM, the downstream gene in a 2-gene operon comprising acnA ([27]; S1 Fig), suppresses the ΔplsX phenotype. For this, a fadM transposon insertion from the Nebraska Transposon Mutant Library, interrupting FadM at amino acid position 5, was transferred by ϕ80-mediated transduction into the JE2 ΔplsX fadM2 suppressor (strain 7.1; S2 Table) [28]. The ΔplsX fadM::Tn strain failed to grow on FA-free medium (Fig 3B). Since total loss of FadM does not promote ∆plsX suppression, we conclude that some FadM activity is required for S. aureus ΔplsX suppression. As FadM I38T, Y90F, and Y133F mutations each map to the acyl-CoA thioesterase binding tunnel, we propose that the FadM ACP binding activity is involved in ∆plsX suppression (Fig 3C).

Impact of ΔplsX suppressors on bacterial growth properties

Growth of ΔplsX suppressors was assessed in RN-R and JE2 backgrounds in non-FA-supplemented BHI medium (Fig 4A). The initial ΔplsX deletion strains show only residual growth, likely due to trace amounts of FA in medium [16], unless supplemented with FAs. In contrast, both fabF and fadM suppressor mutants grew similarly, but their growth phenotypes differed according to the parental strains: in the RN-R background, growth of both tested suppressors was intermediate between the WT strain and the ΔplsX mutant. In contrast, JE2 derivative suppressors grew like the parental strain. A distinguishing feature between the two S. aureus strains is the production of staphyloxanthin pigment by JE2 but not by RN-R (an RN4220 derivative), whose hydrophobic long-chain isoprenyl moiety [29] could conceivably stabilize the membrane in suppressor strains. However, decreased pigmentation of ΔplsX, and both fabF and fadM suppressors compared to WT and fadM::Tn controls (Fig 4B) goes against this hypothesis. Possibly, stress response functions produced in JE2 but not in the 𝜎^B-defective RN-R strain [30] contribute to better growth of suppressors in that background.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. Behavior of fabF and fadM suppressors of∆plsX without exogenous FAs.

A. Growth of WT S. aureus (RN-R or JE2), ΔplsX, and ΔplsX fabF or ΔplsX fadM suppressors in BHI medium. Left: RN-R WT, and derivatives: ΔplsX, and ΔplsX suppressors fabF1 (encoding FabF^A119E) and fadM2 (encoding FadM^Y90F). Right: JE2 WT, and derivatives: ΔplsX, and ΔplsX suppressors fabF1, fadM1, fadM2, and fadM3 (encoding FabF^A119E, FadM^I38T, FadM^Y90F, and FadM^Y133F respectively). Mean and standard deviation of independent triplicate cultures are shown. B. BHI + C18:1-grown overnight cultures of the indicated strains were pelleted, and photographed to compare pigmentation. C18:1 was added to include ΔplsX in comparisons. C. FA profiles of RN-R (left)- and JE2 (right)- derived WT, ∆plsX, and fabF or fadM suppressors determined from BHI-grown cultures. FA assignments are as indicated. ai, anteiso; even numbered FAs are saturated (e.g., C16 is C16:0). The main FA component of WT S. aureus is ai15 [7]. Peak heights correspond to relative responses (mV) of each FA in a sample. The proportions (in %) of C18:0 plus C20:0 of total FAs are indicated over the brackets. S3 Table provides complete FA profile information. D. Acyl-ACP species in S. aureus JE2-derived ΔplsX, and ΔplsX suppressor strains. JE2 WT, ΔplsX fadM1, ΔplsX fadM3, and ΔplsX fabF1 (isolates 7.5, 8.7, and 7.7, respectively; S2 Table) were grown in BHI. The ΔplsX strain was pre-grown in BHI + C18:1 medium, and FA-starved for 2h in BHI prior to harvesting. Extracts were loaded on non-denaturing acrylamide gels, and immunoblotted using anti-ACP [7,36,53]. Longer FA moieties on acyl-ACP migrate faster on gels (wedge; [36,54]). White arrow, accumulated acyl-ACP produced by FASII is not processed to phospholipids [5] (Fig 1). Results are representative of biological duplicates.

https://doi.org/10.1371/journal.pgen.1012165.g004

Membrane FAs are shortened in fabF and fadM suppressors of ΔplsX

We considered that the fabF1 point mutation in the ∆plsX fabF1 suppressor strain might alter FabF processivity, suggesting a shift in FASII products. We also asked whether the mutations in fadM, encoding a possible FabF partner, might have similar effects. We therefore assessed membrane FA composition of both ΔplsX suppressors in RN-R and JE2 backgrounds in medium devoid of FAs (Fig 4C and S3 Table). The fabF suppressors produced membrane FAs of markedly shorter overall length in both strain backgrounds. Notably, straight-chain FAs C18:0 and C20:0, represent 23–29% of total FAs in parental strains but are essentially absent in fabF suppressors. This suggests that processivity of FabF^A119E expressed from the fabF1 suppressor allele is reduced compared to WT FabF. In S. aureus, branched chain precursors are preferred substrates for FabH (see Fig 1; [31]), which might lead to higher proportions of branched chain FAs (ai15 and ai17) in fabF1 suppressors. FA profiles in the fadM2 suppressor showed an ~ 2-fold decrease in proportions of C18:0 and C20:0 and a concomitant increase in shorter chain fatty acids compared to those in the respective parental strains (Fig 4C and S3 Table). These results confirm that both fabF and fadM mutant suppressors of ΔplsX lead to decreased proportions of long chain saturated FAs in phospholipids.

∆plsX suppressors display accrued β-lactam susceptibility

β-lactam resistance of methicillin resistant S. aureus (MRSA) is associated with the presence of penicillin binding protein Pbp2A’ (encoded by mecA), and is a main cause of treatment failure [32,33]. Interestingly, disturbance of membrane lipid microdomains affects Pbp2A’ oligomerization and compromises resistance [34]. We asked whether the membrane changes induced by ∆plsX suppressors, notably their shorter FAs (Fig 4C) might also impact β-lactam antibiotic resistance profiles. For this, amoxicillin minimal inhibitory concentrations (MICs) were measured by Etests. Tests were performed on BHI solid medium without and with 250 µM C18:1; the FA addition allowed us to include ∆plsX in these comparisons (S5 Fig). Control strains (WT, fadM::Tn, and ∆plsX) showed comparable resistance to amoxicillin in both media (0.25-0.38 µg/ml by Etest). Remarkably, all ∆plsX suppressors, carrying fabF or fadM mutant alleles, showed accrued amoxicillin susceptibility (0.064-0.094 µg/ml). It was initially perplexing that FA addition did not restore amoxicillin resistance to levels of the ∆plsX mutant. One might speculate that fabF or fadM suppressor mutations induce stress, e.g., due to increased intracellular FAs, leading to expression changes affecting PBP2A’ folding or stability; alternatively, FA incorporation could be altered in suppressor strains. The mechanisms responsible for this β-lactam-susceptibility phenotype remain to be elucidated.

Acyl-ACP species vary in JE2 WT, ΔplsX, and ΔplsX suppressors

Since PlsX uses acyl-ACP, the FASII end product, to initiate phospholipid production, its deletion provokes acyl-ACP accumulation (Fig 1; [5,35]). Exogenous FA addition resolves accumulated acyl-ACP in a ΔplsX mutant [12]. We assessed acyl-ACP intermediate accumulation in the JE2 ΔplsX mutant and derivative suppressors by non-denaturing polyacrylamide gel electrophoresis, and probing with anti-ACP antibodies [7,36] (Fig 4D). Unlike the WT strain, the ΔplsX mutant was distinguished by the presence of a fast-migrating band corresponding to ACP bound to long-chain FA generated by FASII, likely the dominant FA ai15. The ∆plsX strain assimilates acyl-ACP to form phosphatidic acid (PA) only if an exogenous FA (e.g., C18:1) is added to initiate PA synthesis (Fig 1). The long acyl-ACP species was absent in ΔplsX fabF1 and ΔplsX fadM1, M2, or M3 suppressors. This observation, and good growth of suppressor strains (Fig 4A) indicate that acyl-ACP is processed to complete phospholipid synthesis and bypass the phospholipid synthesis block in ΔplsX.

ΔplsX suppressors do not compensate reverse PlsX activity, which is required to bypass the FASII pathway

The above results show that fabF and/or fadM suppressors compensate the absence of PlsX in normal growth conditions, i.e., in converting acyl-ACP to acyl-PO₄. We asked whether fabF and/or fadM suppressor strains could also complement the reverse PlsX activity, i.e., conversion of acyl-PO₄ to acyl-ACP (Figs 1 and 5A). If this were the case, FASII inhibition in ΔplsX suppressors would be bypassed by FA supplementation to growth medium. To test this, WT strain JE2, and isogenic ΔplsX, ΔplsX fabF1, and ΔplsX fadM2 strains were grown in FA-supplemented medium (SerFA, containing 250µM of an FA mixture and 10% fetal calf serum; see Materials and Methods), without and with the FASII inhibitor AFN-1252 [37] (Fig 5B). All strains grew to equivalent densities in the absence of inhibitor. The JE2 WT strain overcame the presence of AFN-1252 in SerFA [6,8]. In contrast, neither the ΔplsX mutant nor the ΔplsX suppressors could bypass the FASII inhibitor. These findings show that fabF or fadM suppressor mutations do not compensate the PlsX reverse reaction, and that FASII must be active for ΔplsX suppressor growth.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 5. Synthetic lethal effect of a combined FASII and PlsX block in∆plsX and suppressor strains.

A. Expected effects of FASII inhibition on an S. aureus ∆plsX mutant. In a WT strain, FASII inhibition by mutation or by antibiotic action is overcome by exogenous FAs, which are phosphorylated by Fak [4]. FA-PO₄ is the substrate for PlsY and for reverse PlsX activities, and is thus used for phospholipid (PL) synthesis at both the 1 and 2 positions. If both FASII and PlsX are blocked as shown, exogenous FAs can only fill position 1. The FASII defect or PlsX absence can be rescued separately by FA supplementation, but not the combination. Green, exogenous FAs and FA kinase; purple, plsX deletion; grey, pathways and products affected by combined FASII and PlsX block. B. Effects of anti-FASII treatment on JE2 ΔplsX, and ΔplsX suppressor strains in FA-supplemented medium. Strains were precultured in SerFA, and grown in SerFA without or with addition of 0.5 µg/ml anti-FASII AFN-1252. When the anti-FASII is added, WT S. aureus strains incorporate FAs to constitute membrane phospholipids, thus bypassing inhibition [6,8,55]. As ΔplsX suppressors cannot generate FA-ACP needed for FASII bypass [6,9,12], they remain sensitive to anti-FASII. Mean and standard deviation of independent triplicate cultures are shown. Black lines, no anti-FASII, red lines with anti-FASII. fadM2 encodes FadM^Y90F; fabF1 encodes FabF^A119E.

https://doi.org/10.1371/journal.pgen.1012165.g005

As the ΔplsX‍ fabF and fadM suppressors behaved similarly in all the above phenotypic tests (Figs 4,5, S4, and S5), we considered that these alleles might cooperate in a single pathway.

FadM is required for emergence of ΔplsX‍ suppressors mapping to FabF

As fadM interruption abolished ∆plsX suppression (Fig 3), we reasoned that suppressors arising in the ΔplsX fadM::Tn mutant would map to fabF. This hypothesis was tested by comparing suppressor emergence in JE2 ΔplsX and ΔplsX‍ fadM::Tn backgrounds on solid medium without FA supplementation. We also tested for a stimulatory effect of the FabF inhibitor platensimycin as seen for the ΔplsX mutant (Fig 2). Unexpectedly, no suppressors arose after extended (> 1 week) plate incubation at 37°C, compared to frequent mutant emergence in the ΔplsX background after 1–2 days (Fig 6A, upper). Moreover, no stimulatory growth ring was detected surrounding the platensimycin spot (Fig 6A, lower). These results affirm that FadM is needed to obtain FabF suppressors of ΔplsX. Finally, we introduced the fadM::Tn allele into the ∆plsX fabF1 suppressor mutant. Compared to good growth of ∆plsX fabF1, introduction of the fadM::Tn allele arrested growth in the absence of exogenous FA, proving that fadM is required for fabF1 rescue (Fig 6B).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 6. fabF suppressors of ΔplsX require the presence of a fadM allele.

A. FadM is present in JE2 ΔplsX (left) and absent in ΔplsX fadM::Tn (right, constructed from strain 7.1: S2 Table). Strains were plated on BHI solid medium (100 µl of OD₆₀₀ = 0.005), and in the lower set, spotted with platensimycin (P; 1.5 µg). Plates were photographed after 6 days at 37°C. Black arrowheads point to ΔplsX suppressor clones. The dashed curved line indicates the growth ring surrounding platensimycin spotting. The small colonies surrounding platensimycin on ΔplsX fadM::Tn lawns (right) failed to regrow on BHI medium. B. FadM is present in the JE2 ΔplsX fabF1 suppressor strain (left) and absent in ΔplsX fabF1 fadM::Tn (right, constructed from strain 7.8: S2 Table). Strains were streaked on BHI solid medium (upper), or BHI + C18:1 and 5 µg/ml erythromycin (Ery; lower). Plates were photographed after 24 h incubation at 37°C. Results are representative of three independent cultures and platings.

https://doi.org/10.1371/journal.pgen.1012165.g006

The potential relevance of FabF-FadM cooperation was examined in a WT background by comparing platensimycin inhibition zones in the WT and fadM::Tn derivative on BHI medium (S6 Fig). While inhibitory zones were equivalent, the WT strain displayed a distinct growth ring (24 h incubation) that is not formed in the fadM::Tn mutant. The growth ring, indicating rescue by platensimycin, also requires FadM in the WT background. Based on all the above findings, we conclude that FabF and FadM have cooperative roles.

Strains, growth conditions, and plasmids

Strains and plasmids are listed in S4 Table. S. aureus strains are derived from USA300_FPR3757 strain JE2 [28] and RN4220-R (called here RN-R; [15]). RN-R is a readily transformable strain derived from RN4220, of the NCTC 8325 lineage [30]. It was repaired to restore a functional fakB1 gene, which is defective in the entire NCTC 8325 lineage [15]). S. aureus parental strains and ΔplsX-suppressor mutants were grown in BHI medium (Brain Heart Infusion; Gibco, France) or in LB (Lysogeny broth; [40]) as specified, at 37°C with shaking. Medium was supplemented with C18:1 (oleic acid) at a concentration of 250 µM (Larodan, Sweden) when indicated. The FabI inhibitor AFN-1252 (Med ChemExpress, France) and FabF inhibitor platensimycin (Bioaustralis-Tebubio France) were used as specified [18,41].

∆plsX mutant construction

The ΔplsX mutants were constructed in RN-R (RN4220 repaired for a defective fakB1 gene in the NCTC8325 lineage; [15]) and JE2 backgrounds. The preparatory plasmid for plsX deletion was constructed using the Gibson assembly method to insert DNA segments flanking the plsX orf into Sma1-linearized pMAD, a thermosensitive plasmid expressing erythromycin resistance [42,43]. Oligonucleotides for ∆plsX construction are in S5 Table. Constructions were verified by DNA sequencing of PCR-amplified fragments. The double cross-over gene replacements were obtained as described [15,43] in RN-R and JE2 strains. Deletions removed nucleotides 1,228,616-1,229,594 in JE2, and 1,148,670–1,149,648 in RN-R, as per respective annotated genomes (GenBank Nucleotide accession codes NC_007793.1 and NC_007795.1). The resultant strain genotypes were confirmed by whole genome sequencing (Eurofins, Germany).

plsX cloning and complementation of the ∆plsX mutant in RN-R and JE2 strains

To confirm that strain characteristics were due to the plsX deletion, the intact plsX gene was cloned with its natural promoter (upstream of fapR) on multicopy plasmid pIM-locus1 (a pIMAY derivative [44,45]). Cloning was performed by Gibson assembly with oligonucleotide primers (S5 Table).

Isolation of suppressor mutants

RN-R ΔplsX and JE2 ΔplsX strains were first streaked on solid BHI medium containing 250 µM C18:1, an FA that promotes growth. For each strain, a single colony was used to inoculate a liquid culture (in BHI + C18:1). When the culture reached exponential phase, bacteria were centrifuged, resuspended in BHI, and adjusted to OD₆₀₀ = 0.1. Dilutions were plated on solid medium to obtain ~10⁵ to ~10⁷ colony forming units per plate on solid BHI medium without FAs. Colonies were scored after two or more days of incubation at 37°C, and selected at random for restreaking on solid BHI medium. Liquid cultures were prepared in BHI, and stocked in 20% glycerol for storage at -70°C. Selected mutant genotypes were confirmed by whole genome sequencing or by PCR fragment sequencing (Eurofins, Germany) to identify mutated loci.

Genomic DNA preparation for whole genome sequencing

Bacterial cell pellets, prepared from 3 mL liquid exponential phase cultures, were centrifuged at 8,000 rpm for 5 min. Pellets were re-suspended in 800 µL Tris-EDTA buffer (10 mM Tris- HC1, pH 8.0/1 mM EDTA) plus 20 µL lysostaphin 5 mg/mL, and incubated for 45 min at 37°C. DNA was prepared by DNeasy Blood & Tissue Kit (Qiagen, France) and samples were outsourced for whole genome DNA sequencing using 2x150 bp paired end chemistry and bioinformatics analyses (Eurofins Genomics, France) [6]. SNPs that differed in non-antibiotic-treated JE2 and RN-R strains from those in the reference sequences (GenBank Nucleotide accession codes NC_007793.1 and NC_007795.1 respectively) were subtracted prior to variant identification. Variants were identified as representing ≥80% of reads in sequences for which there were at least 100 reads.

PCR amplification and confirmation by DNA sequencing

Selected candidate colonies were suspended in 100 μL of 0.9% NaCl with 1 μL of lysostaphin from a 5 mg/mL stock. Tubes containing suspensions were shaken for 45 min at 37°C. Contents were then transferred to tubes containing glass beads and samples were vortexed for 1 min. After a 30 second centrifugation at 8000 rpm, PCR amplification was performed on 1 µl of supernatant, using primers (S5 Table) and Phusion DNA polymerase (ThermoFisher Scientific, France) according to manufacturer protocol, with a SimpliAmp Thermal Cycler (ThermoFisher Scientific, France). PCR products were analyzed on a 0.8% agarose gel prepared in 40 mM Tris acetate 1 mM EDTA buffer containing 0.5 μg/mL ethidium bromide. Gels were run at 100 V for 30 min in an electrophoresis chamber (Mupid ONE; Dutscher, France), and photographed under UV light using the EBOX VX5 imaging system (Vilber Lourmat, France). PCR samples were purified using the NucleoSpin Gel and PCR Clean-up kit (Macherey-Nagel, France), and sent for DNA sequencing (Eurofins, France).

Effects of the FabF inhibitor platensimycin on ΔplsX and ΔplsX fadM::Tn strains

Strains were grown in BHI + C18:1 overnight, washed once in BHI medium, and resuspended in BHI to a final density of OD₆₀₀ = 0.1. From this, 100 µl of the bacterial suspension was plated on solid BHI medium. Platensimycin (1.5 µg in 3 µl) was spotted in the center of plates, which were then incubated at 37°C for 4–7 days and photographed.

Effect of glucose on growth of ∆plsX suppressors

Strains were precultured in liquid LB (n.b., LB lacks glucose) or LB with C18:1 for ΔplsX at 37°C. After overnight growth, cultures were diluted to OD₆₀₀ = 0.05 and grown for 3–4 hours. ΔplsX cultures were first centrifuged and resuspended in fresh LB medium to remove free FAs prior to regrowth. Bacterial densities were then adjusted to OD₆₀₀ = 1. Cultures were streaked on solid medium, corresponding to LB or LB containing 0.5% glucose, which represses fad operon expression [25,27]. Plates were incubated at 37°C for 24 hours and photographed.

Growth kinetics

Three mL precultures were prepared in BHI for parental and ΔplsX suppressor strains, or in BHI containing 250 µM C18:1 for ΔplsX mutants. After overnight growth, culture densities were adjusted to OD₆₀₀ = 0.05 in 200 µL BHI medium without or with C18:1 (250 µM), in 96-well microtiter plates. Growth was followed on a Spark M10 plate reader (Tecan, France). For experiments in which samples were taken for FA determinations or for Western blot experiments, cultures were prepared in 3–10 ml liquid cultures, and were monitored manually by OD₆₀₀ measurements every two hours. Amoxicillin susceptibility was determined using AMOXICILLIN AC 256 Etests (BioMérieux, France).

Determination of membrane FA profiles

Bacterial cultures (equivalent of OD₆₀₀ = 2) were centrifuged, and pellets were washed once in 1 ml of 0.02% Triton in 0.9% NaCl, then twice with 1 ml of 0.9% NaCl. FA extractions were performed as described [6,46]. Briefly, 0.5 mL of 1N sodium methoxide in methanol was added to bacterial pellets in 2 ml microfuge tubes, and then subjected to a 5 min ultrasound bath. Then 0.2 mL heptane spiked with methyl-10-undecenoate (Sigma-Aldrich) as internal standard was added, followed by 1 min vortexing, and short centrifugation for phase separation. FA methyl esters were recovered in the heptane phase. Analyses were performed in a split-splitless injection mode on an AutoSystem XL Gas Chromatograph (Perkin-Elmer) equipped with a ZB-Wax capillary column (30 m x 0.25 mm x 0.25 μm; Phenomenex, France). Data were recorded and analyzed by TotalChrom Workstation (Perkin-Elmer). S. aureus FA peaks were detected between 12 and 32 min of elution, and identified with retention times of purified esterified FA standards.

Detection of ACP products

Pellets from OD₆₀₀ = 10 equivalent of exponential phase cultures were collected by centrifugation (8000 rpm for 5 min), and then washed twice with PBS (Phosphate Buffer Saline)-Triton X100 0.1%, and once with PBS. Pellets were re-suspended in 600 µL PBS supplemented with 6 µL antiprotease (100X), 50 µg/mL lysostaphin and 50 µg/mL DNAse and incubated with shaking for 25 min at 37°C. Protein concentrations were determined [47], and extracts (10 μg) were loaded onto non-denaturing gels containing 1M urea and 14% acrylamide gel as described [5,7]. After electrophoresis, gels were subject to electroblotting for transfer to a PVDF membrane (Biorad, France) using the 25 V and 1.3 A setting on a power blotter semi-dry transfer system (Invitrogen, France) for 7 min. Primary anti-ACP antibody (Covalabs, France; [7]) was used at 1:500 dilution and secondary anti-rabbit IgG conjugated with peroxidase (Invitrogen, France) at a 1:5000 dilution. ACP-reactive bands were identified using the ECL kit (Perkin Elmer, France) and the fluorescent signal was recorded using ChemiDoc (BioRad, France).

Phage transductions

Transductions of fadM::Tn transposon insertion from JE2 into ΔplsX suppressors fadM2 and fabF1 were performed as described using ϕ80 phage stock [48,49]. The fadM transposon insertion [28] was transduced into JE2 ΔplsX fadM2 suppressor strain 7.1 and into JE2 ΔplsX fabF1 suppressor strain 7.8, giving rise respectively to ΔplsX fadM::Tn and ΔplsX fabF1 fadM::Tn, using erythromycin (5 µg/ml) as selection (S4 Table). Insertions were confirmed by PCR.

FASII bypass of WT and ΔplsX derivative strains

Anti-FASII bypass was evaluated as described [6]. Strains were grown overnight in liquid SerFA medium (SerFA, BHI containing 10% newborn calf serum and an equimolar mixture of 83 µM each C14:0, C16:0, and C18:1 [Larodan, Sweden]). Cultures were then diluted to OD₆₀₀ = 0.05 in SerFA containing or not 0.5 µg/ml of the anti-FASII AFN-1252 (here called AFN; [37]; MedChemExpress, France). Growth was monitored using the Spark plate reader (Tecan, France).

Whole-genome sequencing data of S. aureus ∆plsX mutants and suppressors

Data is available in the Mendeley database: https://doi.org/10.17632/mzw2429w3m.1.