Transcriptomic Responses and Survival Mechanisms of Staphylococci to the Antimicrobial Skin Lipid Sphingosine

ABSTRACT

Sphingosines are antimicrobial lipids that form part of the innate barrier to skin colonization by microbes. Sphingosine deficiencies can result in increased epithelial infections by bacteria including Staphylococcus aureus. Recent studies have focused on the potential use of sphingosine resistance or its potential mechanisms. We used RNA-Seq to identify the common d-sphingosine transcriptomic response of the transient skin colonizer S. aureus and the dominant skin coloniser S. epidermidis. A common d-sphingosine stimulon was identified that included downregulation of the SaeSR two-component system (TCS) regulon and upregulation of both the VraSR TCS and CtsR stress regulons. We show that the PstSCAB phosphate transporter, and VraSR offer intrinsic resistance to d-sphingosine. Further, we demonstrate increased sphingosine resistance in these staphylococci evolves readily through mutations in genes encoding the FarE-FarR efflux/regulator proteins. The ease of selecting mutants with resistance to sphingosine may impact upon staphylococcal colonization of skin where the lipid is present and have implications with topical therapeutic applications.

INTRODUCTION

Human skin is a protective barrier with lipids that contribute a major role in establishing the innate defensive properties. Skin lipids of human epidermis have two distinct origins: sebaceous glands, and keratinocytes of the stratum corneum. Composition of these two lipid fractions differ markedly with sebaceous lipids being predominantly squalene, wax esters and triacyclglycerols, whereas mostly the epidermal lipids comprise equimolar amounts of ceramides, free fatty acids, and cholesterol (1–3). The major antimicrobial lipids of epidermal and sebaceous fractions also differ; both contain antimicrobial fatty acids, but the hydrolysis of ceramides generates antimicrobial sphingosines only in the epidermal lipids fraction.

Sphingosines are unsaturated long chain amino alcohols, and their rapid antimicrobial activity contributes to growth limitation of skin pathogens, such as Staphylococcus aureus (4, 5). S. epidermidis is a dominant skin colonizer but the effects of sphingosines upon their colonization are unknown. Reduced sphingosine concentration was correlated with increased skin colonization by S. aureus in patients with atopic dermatitis (6), and with S. aureus lung infections in cystic fibrosis (CF) patients (7). The concentration of sphingosines is significantly reduced in the lungs of CF patients due to lower ceramidase activity and in CF mouse models (8), which are both more susceptible to pulmonary infections. The increased susceptibility to S. aureus due to the imbalance between ceramides and sphingosines is reversible with sphingosine treatment in mouse models (7). Sphingosine was proposed as a novel treatment against staphylococcal infection; as a pretreatment on implanted medical devices (9, 10), or as an inhaled treatment for pulmonary S. aureus infections in CF patients (7, 11).

Sphingosine was recently shown to interact with the cell membrane via cardiolipin to kill S. aureus through rapid membrane permeabilization (11). The authors speculate that cardiolipin-sphingosine interaction forms rigid domains within the membrane, as described for the aminoglycoside 3′, 6-dinonylneamine (11, 12). Knowledge of resistance mechanisms within staphylococcal communities would help to inform usage and to safeguard against rapid spread of resistance to this novel antimicrobial.

In this study, we compared the gene expression of both S. aureus and S. epidermidis in response to micromolar concentrations of d-sphingosine and identified a common stimulon. Discrete responses were revealed, and a common mode of resistance was identified through experimental evolution of both species of staphylococci.

RESULTS

Staphylococcal growth inhibition by d-sphingosine.

Sphingosines are potently antimicrobial to staphylococci (4, 24), which was confirmed in this study by testing multiple S. aureus and S. epidermidis strains for their d-sphingosine MIC and MBC (Table 1). Comparison of these data revealed an MIC range from 8 to 38 μM with MBC values that typically matched, suggesting that once the concentration of d-sphingosine reached a threshold level its activity was lethal. The MIC and MBC values determined here were used to establish conditions for further examining the responses of staphylococci to sphingosines.

TABLE 1.

d-sphingosine MICs and minimum bactericidal concentrations (MBCs) for S. aureus and S. epidermidis. Data represent a minimum of 3 independent assays

Species and strain	MIC (μM)	MBC (μM)
S. aureus
SH1000	8	8
Newman	16	38
MRSA 252	38	38
MSSA 476	38	38
BL137	38	38
S. epidermidis
Tü3298	38	38
NCTC 1457	8	8
Rp62a	38	38
B19	8	8
O16	8	8

Transcriptomic response to d-sphingosine challenge.

A transcriptomics led approach was taken to investigate the outcomes of S. aureus and S. epidermidis challenge with d-sphingosine and determine whether there were discrete responses. Exponentially growing cultures of S. epidermidis Tü3298 and S. aureus Newman were challenged with d-sphingosine for 20 min prior to harvesting of cells for RNA-Seq, using an experimental design described previously (13, 16). A d-sphingosine challenge concentration of 5 μM was chosen as the minimum concentration that the lipid caused growth inhibition of both species under the culture conditions (Fig. S1 in the supplemental material and data not shown). Controls were treated with the corresponding volume of ethanol solvent, which did not affect growth at the concentration used.

The challenge stimulon of S. aureus Newman comprised 1,331 genes that were differentially expressed (DE) in response to d-sphingosine, of which 666 were downregulated and 665 were upregulated (Fig. 1; Table S4 in the supplemental material). Despite the numerous changes, only 91 genes were DE >log₂ 2-fold, of which 57 were downregulated and 34 were upregulated. Contrastingly, the S. epidermidis Tü3298 stimulon showed a more modest number of DE genes in response to d-sphingosine, with only 129 downregulated and 211 upregulated genes. Of these, only 55 were DE >log₂ 2-fold, of which 11 were downregulated and 44 were upregulated. Of the DE genes from each species, 1,071 had homologs in the other species and in common, 101 were upregulated, and 43 were downregulated (Fig. 1; Table S4 in the supplemental material). Confirmation of DE gene expression from the RNA-Seq was done using qPCR for five selected genes (Fig. S2 in the supplemental material).

FIG 1.

Comparison of up- and downregulated genes of S. aureus and S. epidermidis after d-sphingosine challenge. Genes with homologs in both species are indicated in the central ovals, whereas genes without homologues are shown in the peripheral ovals. Common differentially expressed genes are indicated in overlapping portions of the colored shapes, comprising the d-sphingosine challenge stimulon.

By assigning genes from the whole genome to Clusters of Orthologous Genes (COG) classes, and comparing the number of genes up- and downregulated in these classes compared with the expected number if expression was uniform across the genome, it is possible to find functional groups of genes that are more up- or downregulated than expected (Fig. S3 in the supplemental material). We found that S. aureus and S. epidermidis both had enhanced upregulation or metabolism genes, particularly in amino acid transport and metabolism which included histidine, valine, isoleucine and leucine amino acid biosynthesis genes. S. epidermidis had further enriched upregulation of energy production and conversion and carbohydrate transport and metabolism gene classes. Further investigation revealed these to be primarily in pyruvate metabolism, galactose metabolism and carbohydrate uptake, suggesting S. epidermidis alters metabolism away from glycolysis in response to d-sphingosine challenge while S. aureus does not.

Phosphate transporter upregulation.

Within the common d-sphingosine stimulon of S. aureus and S. epidermidis (Table S5 in the supplemental material), the pstSCAB phosphate transporter operon was differentially expressed in both species with very high expression in S. aureus (log₂FC 3.1–4.1) (Fig. 2). Shared upregulated expression led us to hypothesize that phosphate homeostasis contributes to d-sphingosine resistance or metabolism in staphylococci. The transcription upregulation was confirmed by qPCR for pstB of S. epidermidis Tü3298 and S. aureus Newman, and this upregulation response was conserved across a range of S. aureus strains (Fig. 2).

FIG 2.

Transcription of key genes postchallenge with sphingosine and their role in resistance. (a) Transcription of pstSCABphoU for control untreated (U, blue) and treated cells (T, purple) of S. aureus and S. epidermidis. Read counts were normalized by total reads per library. Representative data of 3 replicates is shown. (b) qPCR of pstB and vraX across different S. aureus and S. epidermidis strains showing fold change after challenge with d-sphingosine, average fold change and standard error for 3 replicates is shown. (c) d-sphingosine MICs of S. aureus wild type and mutant strains with standard error for 3 replicates.

The contribution of the PstSCAB transport system operon to d-sphingosine survival was tested by constructing an allelic replacement mutant of pstS in S. aureus Newman. The pstSCAB::tet allelic replacement mutant exhibited a lower d-sphingosine MIC relative to its parental strain (6.6 vs 16.5 μM) (Fig. 2). A pstS::mariner transposon mutant exhibited a matching MIC as the operon deletion confirming the phenotype. As phoU1 is downstream of the PstSCAB operon and is known to contribute to persister formation, resistance to the membrane-damaging agent SDS, and virulence gene expression in S. aureus (25), we hypothesized this gene may also contribute to d-sphingosine resistance. This was tested by generating Newman phoU1::mariner transposon mutant, which had a lower MIC compared with its isogenic parent (9.9 μM vs 16.5 μM).

A pstSCAB operon complementation strain could not be generated in S. aureus despite multiple strategies. Presumably the membrane protein expression of the operon in a multicopy plasmid was toxic in E. coli. We have observed this phenomenon previously with S. aureus transporter proteins (26, 27) and we were unable to circumvent this issue by cloning the pstSCAB operon directly into S. aureus RN4220, a strategy that we employed previously. However, as the same gene expression changes were seen in multiple strain backgrounds, and the same reduction in resistance was seen in both the pstSCAB::tet and pstS::Tn strains we feel this provides strong evidence for the role of the Pst operon in d-sphingosine resistance.

Altered TCS and virulence gene expression.

RNA-Seq also revealed sphingosine challenge led to downregulation of the SaeSR two-component system (TCS) genes saeS and saeR and upregulation of the VraSR TCS genes vraS and vraR of both S. aureus Newman and S. epidermidis Tü3298. In S. aureus, this was accompanied by downregulation of multiple virulence factor genes encoding adhesins, immunomodulatory factors and toxins (Table S4 in the supplemental material). Many of these factors promote host colonization as well as virulence and are part of the SaeSR regulon (28, 29) suggesting d-sphingosine decreases TCS activation. S. aureus Newman has a point mutation in saeS that results in constitutive activation of the Sae system regulator SaeR (30) and has previously been shown to cause a hyperactive SaeSR response (31). However, as we observe downregulation not upregulation, saeR was confirmed to be downregulated by qPCR (Fig. S2), and a similar response was seen in S. epiderimidis, this supports a genuine downregulation of the SaeSR regulon by d-sphingosine.

Increased expression of vraSR was observed in both S. aureus and S. epidermidis, together with its cell wall stress stimulon in S. aureus (e.g., pbpB, murF, and msrA1; Table S4 in the supplemental material) (32, 33). The capacity of VraSR to regulate the response of S. aureus to d-sphingosine was supported with the reduced MIC of both vraS::tet (6.6 mM) and vraR::tet (6.6 mM) allelic replacement mutants compared with their isogenic parent SH1000 (16.3 μM).

In S. epidermidis, sepA (SETU_02085), encoding a metalloprotease that inhibits LL-37 activity (34), was the only DE virulence-associated gene other than saeS and saeR. While SaeRS TCS enhances virulence in S. epidermidis, its role is primarily linked to anaerobic growth and control of autolysis rather than expression of virulence determinants (35, 36).

Cell wall and membrane gene expression.

There was a pronounced difference between S. epidermidis and S. aureus gene expression for glycerolipid and fatty acid biosynthesis and degradation pathways (Table 2; Table S3 in the supplemental material). In response to d-sphingosine challenge of S. aureus, expression of fatty acid biosynthesis and degradation pathway genes and glycerophospholipid biosynthesis genes were mostly upregulated. Contrastingly, there was little altered gene expression in these pathways with S. epidermidis. Many S. aureus peptidoglycan biosynthesis genes, including murZ, murF, mraY and glycosyltransferase NWMN_1766, were upregulated in contrast to S. epidermidis (Table S4). Increased cell wall and cell membrane biosynthesis is typically associated with cell growth, although the culture data indicated a minor impact following d-sphingosine challenge suggesting instead there could be restructuring of the cell wall and membrane in response to the imposed or consequent stress. Upregulation of the VraSR cell wall stress TCS indicates this TCS could be responsible for the observed for DE of S. aureus cell wall stress regulon genes. The stress impact of d-sphingosine could be evidenced in both staphylococci by upregulation of the gene for the stress response regulator CtsR, and its regulon genes dnaJ and dnaK (Tables S4 and S5 in the supplemental material) (37, 38).

TABLE 2.

Differential gene expression of lipid and fatty acid catabolism and metabolism genes after d-sphingosine challenge^a

^aColoring indicates the extent of log₂ fold DE, with yellow to red indicating low to high upregulation and pale blue to dark blue indicating low to high downregulation, black indicates no homologous gene for that species, white indicates no significant fold change

With respect to altering cell surface features, both S. aureus and S. epidermidis responded to d-sphingosine with DE of cell membrane and cell wall modification genes. In S. epidermidis this included downregulation of mprF and the dlt operon, which encode enzymes that lysinylate peptidoglycan and d-alanylate teichoic acids respectively (39), which could potentially lead to an overall increase of negative surface charge. The S. aureus response reveals upregulated genes of the tag operon, which encodes biosynthesis of wall teichoic acid and its attachment (40). The upregulation of this modification could similarly increase negative cellular charge, particularly if combined with cell wall thickening caused by increased peptidoglycan biosynthesis. At this stage, it is unclear how the amphipathic nature of the sphingosine combined with its positively charged amino group (at neutral pH) relates to these changes.

S. aureus downregulated expression of carotenoid biosynthesis (crtMNOPQ) genes between 1.15 and 1.74 log₂ fold after sphingosine challenge. Modulated pigment expression is a known response to membrane and peroxide stress (41, 42). Carotenoid levels were too low to detect variations at the mid-log growth phase used for RNA-Seq expression data, however surprisingly after 24 h of growth in the presence of 5 μM d-sphingosine there was increased carotenoid production in S. aureus Newman (Fig. S4 in the supplemental material). Consistent with increased carotenoid after overnight culture with sphingosine, cells had a greater survival from 1 mM hydrogen peroxide challenge (Fig. S4). It is therefore unclear if the downregulation of carotenoid biosynthesis is a temporary event or inconsistent with the pigmentation levels at 24 h for other reasons.

Evolution of resistance to d-sphingosine.

While transcriptomic data can be used to identify resistance responses it has limitations, and we supplemented our investigation with an experimental evolution approach to select for S. aureus and S. epidermidis mutants with increased resistance to d-sphingosine.

S. aureus SH1000 and S. epidermidis Rp62a were growth-passaged in concentrations of d-sphingosine ranging from 0.5–128 μM to concentration, with the highest concentrations at which the culture grew being used to seed the next passages. Both species developed stable phenotypes over time that resulted in increased growth in elevated d-sphingosine (Fig. 3). For undetermined reasons, passaged S. aureus had much higher levels of stable resistance to d-sphingosine than S. epidermidis, and had stable increases in resistance more rapidly than S. epidermidis (day 3 vs day 23). The MIC values of S. aureus isolates from days 3 and 5 showed a 2 to16-fold MIC increase (Fig. 3) and although bacteria were passaged for 9 days, there was no additional increase in MIC beyond day 5, indicating a maximum threshold for the conditions tested.

FIG 3.

d-sphingosine MIC of experimentally evolved isolates. d-sphingosine MICs S. aureus (a) and S. epidermidis (b) from passage days 0–9 and 0–23 respectively. For each day of sampling, individual isolates of separate growth passages are labeled A, B, C. Data represent a minimum of 3 independent assays. White bars indicate the representative isolates selected for whole genome sequencing.

S. epidermidis cultures could grow with increased concentrations of d-sphingosine from day 2 when serially passaged (data not shown), but this was not matched with increased MIC when testing evolved isolates. This growth was likely due to adaptive changes or unstable genetic mutations that quickly reverted without the selective pressure of d-sphingosine. By day 23, increased MIC of isolates became fixed with 2 to 4-fold MIC increases. Whole genome sequencing was used to identify mutations contributing to the phenotype of representative isolates of S. aureus with 2, 8 and 16-fold increased MIC, and S. epidermidis isolates with 2- and 4-fold increased MIC. S. epidermidis evolved isolate 21B, which had the same MIC as WT was selected as an internal control, to resolve mutations that did not contribute to increased MIC. Mutations were compared with the sequences from unevolved day 0 isolates to eliminate genomic differences between the parent genome and reference genome.

Genetic variation in evolved isolates.

Genome sequencing of three S. aureus and three S. epidermidis isolates with experimentally evolved resistance to d-sphingosine and their parent strain identified single nucleotide polymorphisms (SNPs) (Fig. 3 and Fig. 4; Table S6 in the supplemental material). Non-synonymous intragenic mutations present in more than one isolate are shown in Fig. 4. For S. aureus this included the glutamyl-tRNA synthetase gltX, hypothetical membrane protein saouhsc_01895 and ATP synthase saouhsc_02341 (atpB).

FIG 4.

Variants and MICs of S. aureus and S. epidermidis experimental evolution isolates. SNPs that occurred in more than one isolate, within a gene and caused non-synonymous changes are shown. Dotted line indicates d-sphingosine MIC of WT.

For S. epidermidis, this included histidine kinase of the two-component system yycG (also known as walK), which regulates cell wall and membrane composition. SNPs in yycG/yycF have previously been shown to provide resistance to cell membrane and cell wall active components in S. aureus (43, 44); however, isolate 21B—which has the same MIC as the day 0 isolate—also had a non-synonymous SNP in yycG (Table S6 in the supplemental material), reducing the likelihood that mutations in yycG increase d-sphingosine resistance. SNPs in the hypothetical protein serp1407, and the ion transporter serp0458 were found in both the S. epidermidis isolates with increased d-sphingosine resistance.

Common to both species with increased d-sphingosine MIC were SNPs in farE and/or farR. In addition, the isolate with the highest d-sphingosine MIC, S. aureus SH1000-5C, had a SNP in farE (FarE-V296A) not shared by any other sequenced isolate (Table S6 in the supplemental material). FarR is a TetR-like regulator of the efflux pump FarE, known to be involved in resistance to another antimicrobial skin lipid, linoleic acid (13, 45). The FarE/FarR regulator-efflux system was therefore considered a strong candidate resistance mechanism to d-sphingosine and was investigated further for its role in sphingosine resistance.

The contribution of FarE/R in d-sphingosine Resistance.

To investigate whether both S. aureus FarE and FarR contribute to d-sphingosine resistance, a farR mutant was constructed and investigated alongside a previously constructed farE mutant (13). Compared with wild-type parent strain MIC (16.3 μM), gene deletion caused reduced MIC: farE (9.9 μM) and farR (11.8 μM) (Fig. 5). Overnight culture of S. aureus with 5 μM d-sphingosine led to adaptations that enhanced survival; untreated S. aureus challenged with subinhibitory concentration (15 μM) d-sphingosine had 42% survival at 15 min, whereas d-sphingosine precultured S. aureus had 69% survival at 15 min (Fig. 5). In contrast, farE and farR mutants showed 0% survival with challenge at 15 μM lipid with or without preculture. This shows that FarE/R are essential for d-sphingosine tolerance, and suggests they play a role in enhanced tolerance through exposure to subinhibitory d-sphingosine.

FIG 5.

Role of FarE/FarR in d-sphingosine resistance. (a) S. aureus SH1000 farR::tet and farE::tet mutant MICs compared to WT. The mean and standard error of 3 replicates is shown. (b) Survival of SH1000 WT, farR::tet and farE::tet in 15 μM d-sphingosine after overnight growth in control conditions (circles) or with 5 μM d-sphingosine (crosses). Representative data from 2 replicates is shown.

DISCUSSION

Staphylococcus species are known for their particular niche success in colonization of skin, an ability that is multifactorial. The various lipids and fatty acids that are produced from epidermal and sebaceous compartments form a key barrier to colonization of skin. Study here of the staphylococcal response to d-sphingosine was made to gain insight of the mechanisms that could overcome activity of this ceramide breakdown product.

Using an RNA-Seq approach, it was determined that both S. aureus and S. epidermidis responded to d-sphingosine challenge using several homologous response networks. Upregulation of VraSR and CtsR and downregulation of SaeSR regulons appear within these data reflecting a stress stimulus response to cell surface challenge. Similar to the effects of linoleic and sapienic acids that affect SaeSR regulation (16, 27), d-sphingosine may contribute to modulating S. aureus colonization factors.

Upregulation of both VraSR and CtsR regulons could map to the mode of action of sphingosines. Sphingosine was recently shown to interact with the cell membrane via cardiolipin to kill S. aureus through rapid membrane permeabilization (11). The authors speculate this could be caused by the cardiolipin-sphingosine interaction forming rigid domains within the membrane, as is indicated for the aminoglycoside 3′, 6-dinonylneamine (11, 12). Both S. aureus and S. epidermidis respond to d-sphingosine by modulating lipid biosynthesis and fatty acid degradation pathways in dissimilar expression patterns, but there was no alteration to cardiolipin biosynthesis gene transcription, with neither cls1 nor cls2 genes being DE.

A common response to sphingosine in both staphylococci was upregulation of the transport operon pstSCAB. The encoded phosphate-specific transport system is a high-affinity, low-velocity, free-Pi transporter system that is structurally similar to ABC transporters (46). Downstream of the pstSCAB operon, the phoU1 gene encodes a transporter-associated protein. PhoU proteins contribute to multidrug tolerance and persistence in S. aureus and impacts on widespread processes beyond inorganic phosphate (Pi) transport (47). Upregulation of proteins encoded by the pstSCAB and phoU operons were also found in antimicrobial peptide ranalexin resistance in methicillin-resistant S. aureus (48). Genetic disruption of the Pi transporter, PstC, had no effect on ranalexin sensitivity, suggesting that MRSA adopts a PhoU-mediated persister phenotype to acquire antimicrobial tolerance, and upregulation of the Pi transporter is not a major component of this response (48). However, inactivation of phoU2 did lead to upregulation of phosphate transport operon expression (25). Inactivation of the pstSCAB operon in S. aureus was shown here to reduce the sphingosine MIC, upregulated phosphate transport could allow staphylococci to incorporate sphingosine into phospholipid biosynthesis, but this was not tested or explored further.

To further identify mechanisms of survival from sphingosine, we employed experimental evolution in the presence of sphingosine. Genome sequencing of isolates with greater MIC to sphingosine revealed patterns of SNPs common to both S. aureus and S. epidermidis that highlighted the FarE-FarR transporter-regulator proteins could contribute to survival. S. aureus FarE was previously identified as a lipid transporter that provides enhanced tolerance to the antimicrobial lipid linoleic acid (13, 45). S. aureus FarR regulates FarE in S. aureus (45, 49) and the S. epidermidis gene for the FarR homologue SERP0247 and efflux transporter gene SERP0245 share synteny. Inactivation of farE or farR in S. aureus reduced the d-sphingosine MIC. Examination of the sphingosine challenge transcriptome shows only the lipid A and amphipathic drug ABC transporter MsbA (50) as a potential exclusion transport mechanism that was upregulated; its role was not tested in the present study and was not identified in the experimental evolution approach. The lack of upregulation of farE or farR in the transcriptomic data is at odds with the enhanced survival of SH1000 to d-sphingosine following overnight culture with sub-MIC sphingosine only in the WT and not in FarE or FarR mutants. This may indicate that d-sphingosine is a weak ligand for FarR activation, and the observed SNPs may either increase the ability of FarR to interact with d-sphingosine or cause constitutively activation of FarR. Alternatively, this could reflect regulatory differences between S. aureus Newman and S. aureus SH1000.

FarR is a member of the TetR family of repressors that form homodimers. Binding of the substrate alters the structure of TetR proteins, reducing their affinity for the DNA binding site and relieving repression of the target gene (51). Most of the S. aureus and S. epidermidis FarR mutations are predicted to be located at the entrance to and within a pocket of the substrate-binding domain. Therefore, the identified changes could either affect how this repressor interacts with its substrate or cause the protein to be in the conformation with low DNA affinity independent of the substrate. A further S. aureus mutation, yielding FarR.A40S, occurred in the α-2 region of the DNA binding domain that may alter DNA binding. The transcriptional effects of a C116R farR mutation were identified in a study of experimentally evolved resistance to rhodomyrtone, a plant acylphloroglucinol antimicrobial, with upregulated FarE expression as the resistance mechanism (49). Of note, this study revealed that the selected mutant had greater virulence in murine models, which demands further study of sphingosine selected mutants.

The FarE transporter was not a regulated exclusion mechanism in response to sphingosine and the mutations require future investigation to identify their exact outcomes. Cellular resistance gained through SNPs in the farE efflux transporter gene could alter affinity (K_m) for d-sphingosine and/or flux (V_max). The FarE transporter is a member of the RND protein family that encode secondary multidrug transporters for a wide range of compounds including drugs, metals, solvents, fatty acids, detergents and dyes (52, 53). The described RND proteins are Mmpl-like transporters and some are described to act as “membrane hoovers” that efflux toxic substrates from the membrane (54). Transmembrane domains of the E. coli multimer AcrB appear to interact loosely, creating vestibules allowing phospholipids and other membrane components access to the central pore which is used for substrate transportation (55). The creation of vestibules through loose interactions of transmembrane domains could explain how substrates in the membrane enter the pump. RND family proteins are characterized by their structure of 12 transmembrane domains, with the first and second, seventh and eighth being linked by extracytoplasmic loops, which likely confer substrate specificity (52, 56).

Both mutant staphylococcal FarE proteins (SERP0245 and SAOUHSC_02866) in this study would bear a single amino acid change facing outwards in an α-helix within the portion of the twelfth and sixth transmembrane domains (respectively) nearest to the extracellular domains. These domains are proposed to have a conserved sequence (53) and may contribute to substrate interaction. Indeed, Guay et al. (57) identified amino acid changes in transmembrane helices of the RND transporter tetA(B) that increased resistance to 9-(dimethylglycylamido)minocycline while decreasing tetracycline resistance. This change in resistance was likely due to differences in substrate specificity (57) and was predicted to occur at similar depth within the membrane to the changes observed here, though in different helices.

Human skin colonization by bacterial genera requires investigation to identify their evolved niche specializations. Such studies will unlock the interplay with the host and reveal balances between commensalism and pathogenesis that could also identify causes of microbiome population changes described as dysbiosis. Concluding from this study of sphingosine survival with these two clinically important staphylococci, the administration of sphingosines to ameliorate topical skin diseases could result in high-level resistance through selection of mutations in genes encoding the FarE-FarR lipid efflux system.

MATERIALS AND METHODS

Bacterial strains and culture.

Staphylococci used in this study are provided in Table S1 in the supplemental material and included the S. aureus strains, SH1000, Newman, SF8300, MRSA252, and S. epidermidis, Tü3298 and Rp62a. Construction of a farE::tet strain of S. aureus SH1000 and its complementation was described in Kenny et al. (13) as SAR2632::tet. Mutants vraS::tet and vraR::tet were constructed previously (13). All bacteria were cultured in Todd-Hewitt broth (THB) unless otherwise stated.

An operon deletion mutant of pstSCAB::tet was generated in S. aureus Newman using pMutin4 by cloning partial pstS and pstB gene fragments together with the tetracycline resistance cassette, as described in Kenny et al. (13). Mariner transposon mutants of pstS, phoR and phoU from the Nebraska Transposon library of S. aureus USA300 were transduced to strain Newman, as described previously (14, 15), using gene-specific primers paired with described upstream/buster primers to confirm orientation and verify their correct allelic replacement; gene-specific primers were used for mutants of phoU, phoR THB and pstS (Table S2 in the supplemental material). d-sphingosine (Sigma-Aldrich) stock solution was prepared at a concentration of 5 mg ml⁻¹ in ethanol. Antibiotics tetracycline, erythromycin, lincomycin and ampicillin were used at concentrations 5 μg ml⁻¹, 5 μg ml⁻¹, 25 μg ml⁻¹, and 100 μg ml⁻¹, respectively.

MIC assay.

MIC assay used a broth microdilution method with overnight broth cultures diluted to provide inoculation at OD₆₀₀ 0.1. MIC was assessed by optical density after 24 h static incubation at 37°C. Minimum bactericidal concentration (MBC) were assessed by viable counting at 24 h.

Sphingosine challenge for RNA analysis.

Bacteria were grown to OD₆₀₀ 0.5 and challenged with 5 μM d-sphingosine for 20 min before harvesting by centrifugation as described previously (16). Three treated and untreated biological replicates were used for RNA-Seq.

RNA and DNA sequencing and bioinformatic analysis.

RNA extraction used enzymatic lysis as described previously (16), RNA integrity was assessed by bioanalyser and only samples with RIN above 7 were used for sequencing. RNA was rRNA depleted using RiboZero rRNA removal kit (Gram-Positive bacteria) and libraries constructed using ScriptSeq (Epicentre). Samples were sequenced by paired-end sequencing on the HiSeq platform (Illumina). RNA from the same samples that was not used for sequencing was converted to cDNA and used for qPCR assays.

Bowtie 2 (17) was used to map reads, read counts were quantified with HTSeq (18), and Edge R (19, 20) was used to determine differentially expressed genes from read counts. Trimmed mean of M values (TMM) was used for normalization, and a generalized linear model (GLM) was used to fit the data to determine differential expression. Benjamin and Hochberg analysis false discovery rate <0.05 was used as the cut off for differentially expressed genes between control and test conditions. S. aureus Newman and S. epidermidis Tü3298 homologous genes were determined by BLAST reciprocal best hit. RNA sequencing and analysis was undertaken by the Centre for Genomic Research (Liverpool, UK).

For COG analysis, protein sequences for all genes within S. aureus Newman and S. epidermidis Tü3298 were submitted to WebMGA function annotation (COG) (21). DE gene lists were labeled with their COG classes and used to calculate the number of DE genes per COG class compared to the whole genome.

Experimental evolution was performed by serial passage, as described previously (22), in broth containing doubling dilutions of sphingosine. For selection of S. aureus and S. epidermidis the maximal assay concentration of the lipid was 128 μM. Control selection experiments were performed in parallel. Experiments were initiated with inoculation of bacteria to OD₆₀₀ = 0.2 for the first passage and cultures were incubated at 37°C. Bacteria growing at the highest concentration of sphingosine after 24 h were passaged forward. For the sequencing of experimentally evolved isolates, DNA was extracted, sequenced, and analyzed as described previously (22), with the exception that all isolates were sequenced individually rather than pooled. Mapping statistics are included as Table S3 in the supplemental material.

cDNA generation and qPCR conditions.

cDNA was synthesized from high integrity RNA samples using the tetro cDNA synthesis kit (Bioline). The SensiFAST SYBR Hi-ROX kit (Bioline) was used for qPCRs. The reaction mix contained 10 μl Sensifast, 0.5 μM each primer, 100 μg cDNA wth DEPC-treated water in a reaction volume of 20 μl. StepOnePlus real-time PCR system (Applied Biosystems) was used for qPCRs. Primer efficiency was confirmed to be within 90–100% as described previously (23). Fold changes were calculated using the standard curve method. At least three biological replicates were used. Primers used for qPCR are listed in Table S2 in the supplemental material.

Staphyloxanthin extraction assay.

Bacteria were grown to OD₆₀₀ 0.5 and challenged with 5 μM sphingosine as for the RNA-Seq experiment. At 24 h cells were adjusted to OD₆₀₀ 5, and 10 ml harvested by centrifugation then resuspended in 1 ml methanol and incubated for 15 min at 37°C with shaking. The suspension was centrifuged for 15 min at 5,000 rpm and using 250 μl of supernatant, absorbance was assayed in triplicate for across wavelengths 350–550 nm.

Sphingosine and hydrogen peroxide challenge survival assays.

Bacteria were cultured with shaking overnight at 37°C in the absence or presence of 5 μM d-sphingosine or with equivalent concentrations of ethanol solvent. Bacteria were harvested by centrifugation and resuspended twice in PBS to wash cells prior to adjustment to OD₆₀₀ 0.1 (∼10⁸ CFU ml⁻¹) and bacteria were challenged with 15 μM d-sphingosine or 7.5 mM hydrogen peroxide. Bacteria were sampled at 0, 15, 30, 60, and 90 min for the sphingosine survival assay and at 0, 15, 30, 60, and 120 min for the hydrogen peroxide survival assay. Survival was assessed by viable counts.

Sequence data accession.

RNA-seq data has been deposited in the ArrayExpress database at EMBL-EBI (www.ebi.ac.uk/arrayexpress) under accession number E-MTAB-9272. Sequence reads of experimentally evolved isolates have been deposited in the European nucleotide archive (ENA) at EMBL-EBI under accession number PRJEB39049.