Two transporters cooperate to secrete amphipathic peptides from the cytoplasmic and membranous milieus

Significance

Organisms compete by secreting cytotoxic peptides and peptidic molecules. These cytotoxins are frequently amphipathic to traverse aqueous environments and reach targets within or across cell membranes. However, this poses a challenge during biosynthesis because newly synthesized compounds can spontaneously partition into the producer’s cell membrane before or after secretion. As a model, we studied the PSM family of amphipathic peptides that are produced by pathogenic Staphylococcus aureus and have provided evidence that two transporter systems cooperate non-redundantly to secrete PSMs from both the aqueous cytoplasm and the hydrophobic membrane. Notably, many biosynthetic gene clusters of antimicrobial bacteriocins encode multiple transport systems, suggesting that this may be a general strategy to boost production and secretion of amphipathic bioactive molecules.

Abstract

Diverse organisms secrete amphipathic biomolecules for competitive gains. However, how cells cope with producing these membrane-permeabilizing molecules is unclear. We focused on the PSM family of secreted amphipathic peptides in the pathogen Staphylococcus aureus that uses two ABC transporters, PmtCD and AbcA, to export peptides across the bacterial cell membrane. We found that increased peptide hydrophobicity favors PSM secretion through PmtCD over AbcA and that only PmtCD protected cells against amphipathic peptides. We propose a two-system model in which PmtCD and AbcA independently export PSMs from either membrane or cytosolic environments, respectively. Our model provides a rationale for the encoding of multiple transport systems on diverse biosynthetic gene clusters used to produce distinct amphipathic molecules. In addition, our data serve as a guide for selectively blocking PSM secretion to achieve antimicrobial or antivirulence approaches and to disrupt established roles of PSM-mediated virulence.

Microbes secrete a chemically diverse array of diffusible small molecules, including peptides, to alter the behavior of, disrupt, or kill intended targets. Many of these molecules are amphipathic to balance the needs of diffusing through aqueous environments and reaching molecular targets either within the hydrophobic membrane or across the membrane and within intracellular compartments. While intense research has focused on the activity and biogenesis of amphipathic molecules, understanding how these molecules are secreted by producing cells is comparatively understudied. In part, this is due to the technical challenges of studying the membrane proteins that conduct export and the low solubility and tendency for aggregation of amphipathic molecules.

Interestingly, distinct transport systems are encoded on biosynthetic gene clusters (BGCs) that encode diverse bacteriocins, which are ribosomally produced and posttranslationally modified peptides that have antimicrobial activity (1–6). Because these bacteriocins are generally toxic to the producing cell, a self-immunity role is frequently attributed to one of the transport systems, whereas a primary secretion role is attributed to the other. However, evidence that clarifies the functional differences that underpin these distinct roles is lacking.

We focused on the cell-based secretion of the phenol-solbule modulin (PSM) family of amphipathic peptides produced by Staphylococci, including the major human pathogen and frequently antibiotic-resistant bacterium Staphylococcus aureus (7, 8). Studying PSM secretion offers rare advantages to interrogate secretion strategies directly from cells. The biosynthesis and the biophysical properties of PSMs from S. aureus are well characterized, and several diverse sets of PSMs are naturally produced across staphylococcal species that enable uncovering associations between PSM secretion and peptide biophysical properties (7, 9–12).

Individual PSM peptides are categorized based on size and split between α- and β-types of 20 to 26 or approximately 45 amino acids in length, respectively. To simplify interpretation, we restricted our analyses to the α-type PSMs because they form single amphipathic helices in solution that otherwise lack substantial tertiary folding (13, 14). Moreover, α-type PSMs are responsible for virtually all known virulence and physiological roles currently attributed to PSMs and are the predominant type produced by S. aureus (8, 15). Notably, the cytolytic activities of PSMs depend on diffusion through aqueous environments and partitioning into cellular membranes, which is facilitated by the ability of α-type PSMs to adopt amphipathic α-helical structures throughout their length.

Because of their amphipathic nature, newly synthesized PSMs before export can spontaneously partition between the cytoplasm and the inner leaflet of the cellular membrane, with which PSMs likely interact interfacially and parallel to the plane of the membrane (16). Two distinct ABC transporters, AbcA and PmtCD, that secrete α-type PSMs have been identified (17–19). Intriguingly, only the pmtABCD operon (which also encodes for the PmtAB transporter that does not transport α-type PSMs) and not the abcA gene provide self-immunity against α-type PSMs, which indicates that export via each transporter occurs through distinct mechanisms and that each transporter may perform specialized roles in secreting PSMs. Here, we provide in vivo cell-based evidence that PSMs access PmtCD and AbcA from discrete environments: aqueous PSMs are efficiently exported by AbcA, whereas membranous PSMs enter PmtCD laterally from the lipid bilayer.

Results

PSM-Specific Transport through PmtCD and AbcA.

We began comparing PSM secretion between PmtCD and AbcA by applying our cell-based secretion assay in which we detect PSMs present in culture supernatants using high-performance liquid chromatography (HPLC) with an inline mass spectrometer (HPLC-MS). We engineered a methicillin-resistant S. aureus (MRSA) USA300 strain LAC devoid of PSMs and in which we deleted the abcA and pmtABCD genes (Δpsmαβhld, ΔpmtABCD, ΔabcA). We transformed this strain with two plasmids: one encoding all five S. aureus α-type PSMs under a xylose-inducible promoter and the other encoding either pmtCD or abcA (Fig. 1A). We then validated that our HPLC-MS detection assay is quantitative. We first revealed a large linear dynamic range dependent on PSM concentration using synthetic PSMs to generate standard curves (Fig. 1B), and then we confirmed that detection of each PSM is identical when spiked into culture filtrates derived from cells expressing pmtCD or abcA (Fig. 1C and SI Appendix, Fig. S1). Finally, because PSMs are toxic to cells lacking the pmtABCD operon (17), we restricted PSM induction to 1 h, which did not noticeably impact cell viability (Fig. 1D). Taken together, we developed a quantitative assay that enabled us to accurately measure PSM secretion by cells through each transporter.

Fig. 1.

Cell-based PSM secretion assay validation. (A) Assay schematic using MRSA LAC Δpsmαβhld, ΔpmtABCD, and ΔabcA transformed with plasmids encoding ABC transporters and α-type PSMs under control of inducible P_xylA promoter. (B) Standard curves of HPLC-MS detection generated with known amounts of synthetic PSM peptides. (C) Indistinguishable (ANOVA P value = 0.28 between pmtCD and abcA groups) detection of synthetic PSMs spiked (50 nM each PSM) into culture supernatants (extracellular fraction) from cells encoding either pmtCD or abcA, demonstrating no differences in PSM elution or sample ionization during HPLC-MS analysis due to distinct sampling environments (i.e., the culture supernatants). (D) Viable cell counts after 1 h of xylose induction demonstrating minimal impact on cell viability (n = 3 per group, mean ± SEM, ANOVA used for significance testing across groups).

In cells that expressed all five α-type PSMs, PmtCD and AbcA differentially transported the individual PSMs: PSMα4 was exported primarily by PmtCD; PSMα1 was exported at similar levels by PmtCD and AbcA; and PSMα2, PSMα3, and δ-toxin were preferentially exported by AbcA. Thus, our data demonstrate substrate-specific export through each PSM transporter (Fig. 2A).

Fig. 2.

Hydrophobicity is a determinant of PSM secretion specificity through S. aureus ABC transporters. (A) PSM secretion after a 1 h of co-induction with all α-type PSMs by cells encoding pmtCD or abcA. (B) PSM secretion after a 1 h induction of individual α-type PSMs by cells encoding pmtCD or abcA (n = 9 per group). (C) Logistic regression showing a significant (P = 0.002) relationship between PSM hydrophobicity and secretion of PSMs by cells encoding pmtCD vs. abcA from data in A. (D) Model of PSM secretion through PmtCD and AbcA. PSMs in the cytoplasm can be exported by AbcA or can partition into the inner leaflet of the cell membrane and be exported by PmtCD. Partitioning of PSMs into the membrane is dependent on hydrophobicity (GRAVY score) and is presumed to be rate-limiting, resulting in the export by PmtCD appearing slower than by AbcA. The hydrophilic and hydrophobic faces of the PSM peptide are colored blue and black, respectively. (n = 9 total of n = 3 per three independent experiments, mean ± SEM nested within independent experiments).

Hydrophobicity is a Determinant of Export Specificity.

Surprisingly however, in cells that expressed each PSM individually, AbcA exported each more efficiently than PmtCD (Fig. 2B). Importantly, that PmtCD and AbcA can export all PSMs regardless of sequence suggests that a determinant other than sequence-specific interactions between individual PSMs with PmtCD or AbcA exists and is needed to explain the specificity we observed when PSMs were expressed as a cohort. We noted that in the latter case, PmtCD and AbcA, respectively, dominated in export of the most (PSMα4) and least (δ-toxin) hydrophobic PSMs as measured by the grand average of hydropathy score (GRAVY), which captures the average residue hydrophobicity over a peptide sequence. To quantitatively compare secretion of each PSM through PmtCD versus AbcA, we calculated a transport preference metric (Materials and Methods) based on the relative amount that each PSM is exported through either PmtCD or AbcA. This enabled us to use logistic regression to statistically test the contribution of hydrophobicity to the transport preference of each PSM. Indeed, we found that the GRAVY score explained 58% (marginal R²) of the transport preference (P = 0.0007, Fig. 2C), demonstrating that hydrophobicity is strongly associated with substrate specificity. Interestingly, the hydrophobic moment, the other primary biophysical parameter of amphipathic helices, was not associated with transport preference (P = 0.62). This is likely due to the narrow range of hydrophobic moments exhibited by S. aureus α-type PSMs (0.404 to 0.599).

While increased hydrophobic interactions between PSMs and PmtCD as compared with AbcA could in part explain why hydrophobicity drives export through PmtCD, we became intrigued by the possibility of an additional mechanism. As amphipathic peptides, PSMs spontaneously partition between aqueous and membranous phases, with hydrophobicity primarily governing the extent of membrane partitioning (20). ABC exporters have evolved to take up substrates from either aqueous or membranous milieus (21, 22). Thus, we hypothesized that PmtCD and AbcA have evolved to cooperatively export PSMs from the membrane and cytoplasm, respectively (Fig. 2D).

We took advantage of the fact that although PSMs are highly conserved within staphylococcal species, PSMs are diverse across the genus, both in number and sequence (SI Appendix, Fig. S2) (8, 9, 23). In contrast, all PSMs known to-date harbor pronounced amphiphilicity and therefore partitioning between lipidic and aqueous environments. Thus, PSM amphiphilicity and not peptide sequence would have constrained the evolution of PSM transporter orthologs across the Staphylococcus genus. This offered a direct test of our model: we predicted PmtCD and AbcA transporter orthologs would not only be capable of exporting S. aureus α-type PSMs but would retain hydrophobicity as a specificity determinant. We cloned pmtCD and abcA orthologs from Staphylococcus epidermidis and Staphylcoccus haemolyticus, two species with characterized α-type PSMs that are dissimilar to S. aureus PSMs, and tested them in our secretion assay using our engineered S. aureus LAC strain expressing S. aureus PSMs. Indeed, all transporters promoted PSM secretion (Fig. 3 A and B), and for each transporter pair tested, we found a significant positive relationship between PSM hydrophobicity and export through PmtCD relative to AbcA (Fig. 3 C and D).

Fig. 3.

Hydrophobicity partially explains PSM export specificity. (A and B) S. aureus PSM secretion in S. aureus LAC Δpsmαβhld Δpmt ΔabcA cells encoding orthologous transporters from S. epidermidis (A) and S. haemolyticus (B). (C and D) Logistic regression showing a significant relationship between PSM hydrophobicity and PSM secretion by cells encoding pmtCD vs. abcA orthologs from data in A and B (C, n = 9 or D, n = 6 for three and two independent experiments for S. epidermidis and S. haemolyticus, respectively; mean ± SEM nested within independent experiments).

Immunity and Viability Roles of PmtCD Involve Membrane Interactions.

We previously reported that chromosomal deletion of the entire pmtABCD operon, which additionally encodes for the distinct PmtAB transporter that has no role in α-type PSM secretion (19), is lethal in a wild-type background due to accumulation of α-type PSMs in the cellular membrane that led to membrane defects (17). However, pmtABCD deletion mutants were readily obtained in a PSM null background. Given our data that PmtCD primarily exports PSMα4—the most hydrophobic PSM—we tested whether deletion of only psmα4 is sufficient to permit deletion of the pmtABCD operon. Indeed, we succeeded in generating a LAC Δpsmα4 ΔpmtABCD mutant (Fig. 4 A and B). Moreover, we also detected δ-toxin secretion in this mutant (SI Appendix, Fig. S3), consistent with our secretion data showing that AbcA primarily transports δ-toxin.

Fig. 4.

Membrane perturbations independently implicate PmtCD in peptide export from the cell membrane. (A and B) PCR product size verification of serial chromosomal deletions of psmα4 (A) and the pmtABCD operon (B) red arrows indicate locations of primer binding for PCR verification. (C) Cell viability of LAC Δpsmαβhld ΔpmtABCD ΔabcA encoding either the pmtABCD operon or the abcA gene on the pRB473 plasmid. K41M, K39M, and K374M are ATPase catalytic mutations that abolish transport. (D) Cell viability of LAC Δpsmαβhld Δpmt cells encoding either the pmtABCD operon or the single transporters, pmtAB or pmtCD (n = 5 per group, mean ± SEM, *P < 0.05, **P < 0.01 by Tukey’s Honest Significant Difference test in comparison with positive control).

Our recent discovery that the pmtABCD operon confers resistance to mammalian antimicrobial peptides (AMPs), such as the cationic amphipathic peptide LL-37 (24), substantiates membranous PSM export by PmtCD. By repeating the in vitro assay, we again observed that the pmtABCD operon conferred resistance to treatment with LL-37. In contrast, abcA had no such effect (Fig. 4C). To determine which Pmt transporter is responsible for protection against LL-37, we tested each transporter individually. Similar to α-type PSM export, we found that pmtCD is responsible for protection against LL-37 with no detectable contribution from pmtAB (Fig. 4D). Taken together, our data demonstrate that PmtCD, but not AbcA confer protection against self-produced PSMs and exogenous AMPs. Because the cytotoxicity of both is primarily caused by membrane perturbation, these results further support our model of PSM export by PmtCD through the membrane environment.

Discussion

Secretion of amphipathic peptides is widespread throughout diverse organisms. How cells evolved to address the secretion of these biomolecules that spontaneously distribute between aqueous and membrane phases remains poorly understood. This is in large part because studying these processes is challenging due to low solubility of the secreted molecule and the membranous transport systems. Circumventing these challenges, we focused on cell-based secretion of PSMs, a family of amphipathic peptides that play diverse roles in staphylococcal physiology and virulence (8). With several unique PSMs coexpressed within S. aureus and other diverse sets produced across staphylococcal species, this system enabled us to make inferences based on the natural differences between PSMs. Our primary discovery is that the hydrophobicity of S. aureus PSMs is strongly associated with secretion through PmtCD. This held true when we tested orthologous transporters from distinct staphylococcal species that encode native PSMs dissimilar to S. aureus PSMs.

Our model (Fig. 2D) provides an evolutionary rationale for having two transport systems of amphipathic biomolecules. Immediately after biosynthesis, PSMs can be exported directly from the cytoplasm through AbcA. The PSMs that are not exported by AbcA can instead partition into the inner leaflet of the cell membrane, the extent of which is governed by the overall hydrophobicity. Notably, our model explains the surprising finding that all α-type PSMs when expressed individually are more likely to be exported by AbcA than PmtCD because an extra step of partitioning into the cellular membrane would be required for export through PmtCD. In addition, this entails that partitioning is rate-limiting for PmtCD export. Importantly, the multidrug-resistant ABC transporter P-glycoprotein, which has been thoroughly characterized as accepting hydrophobic substrates laterally through the membrane (25, 26), provides an instructive precedent: Membrane partitioning of a wide range of substrates was found to be rate-limiting, and hydrophobicity was correlated with ATPase activity (27). Our model also explains why PmtCD, but not AbcA, is essential for staphylococcal viability while PSMs are actively produced because only PmtCD can remove PSMs that partition into the bacterial membrane and prevent PSMs from accumulating to toxic levels.

Several characterized BGCs of amphipathic bacteriocins such as the lanthipeptides nisin, gallidermin/epidermin, and subtilin; the thiopeptide lactocillin; and the cyclic peptides AS-48 and circularin A; and the fibupeptide lugdunin each encode for two transport systems (1–6). Often, secretion is attributed to one system and self-immunity to the other. These attributions are consistent with our model and results reported here. Furthermore, our model predicts that the secretion and self-immunity transport systems encoded in these BGCs export from the cytoplasmic and membranous milieus, respectively. In addition, the self-immunity transporters may enable increased production of the bacteriocin by providing a secondary transport pathway that protects the membrane of the producing cell. Likewise, it is worth noting that PSMs are abundantly produced by Staphylococci (15), which—especially considering its PSM-dependent essentiality—appears to have been enabled by the export role of PmtCD.

While it is often assumed that the immunity transporters remove bacteriocins from the membrane, evidence for this hypothesis has remained limited to peptide release assays (2, 28) from cells due to the difficulty of measuring transport in biochemical assays or in reconstructing the pathway of a substrate from structures. Here we took advantage of the natural diversity of PSMs produced by Staphylococci and provide an independent line of supporting evidence that the immunity transporters accept molecules from the membrane for efflux. As a corollary, our model implies that the non-immunity transporters may accept their substrates directly from the cytoplasm.

Our results are consistent with a high-resolution structure of PmtCD in the apo-state (19). PmtCD adopts a type V ABC transporter fold, which has been associated with lateral access of substrates from the membrane (29, 30). However, this observation relied on a limited set of currently available structures of type V ABC transporters that transport lipids or lipidated molecules residing primarily within the membrane with little to no aqueous presence (31–35). Thus, our data provide functional evidence for membrane access into a type V ABC transporter in which the transported molecules—the PSMs—partition between membrane and cytoplasmic milieus. Interestingly, the ATP-bound state of PmtCD possesses lateral surface grooves lined with hydrophobic residues that span the lipid core of the membrane and are large enough to accommodate α-type PSM peptides. These grooves are plausible sites of initial interaction between PSMs and PmtCD that adhere to our model. Interestingly, PmtCD also protects bacterial cells from host-produced amphipathic peptides such as LL-37. This raises the question of how PmtCD is active against amphipathic peptides originating from outside bacterial cells that would initially partition into the outer leaflet. Amphipathic peptides can spontaneously translocate to the inner leaflet where they could enter PmtCD and be extruded from the membrane. Alternatively, the PmtCD lateral grooves could conceivably accept substrates and therefore extrude amphipathic peptides from the outer membrane leaflet as well. Future functional and structural studies will be needed to clarify the transport trajectory of PSMs and AMPs through PmtCD.

No structure has been reported for AbcA, and the existing literature, aside for PSM secretion, provides no direct evidence of AbcA transport with which we can evaluate our model. One report describes an abcA mutant that exhibits faster autolysis (36), and several describe conflicting results in which abcA inactivation or overexpression increased resistance to β-lactam antibiotics (36–38). Thus, future biophysical and biochemical studies are needed to further elucidate the transport pathway, for PSMs or otherwise, via AbcA.

PSMα2 consistently deviated in our regression analyses. Interestingly, PSMα2 differs from PSMα1 at only three amino acid residues. While specific PSMα2 interactions with AbcA could help explain the deviation, we also note that PSMα2 exhibits substantially less α-helical structure in aqueous buffer than PSMα1 (39). Thus, because peptide folding impacts membrane partitioning (20), PSMα2 may partition less into the cellular membrane than would be predicted by hydrophobicity alone, which in turn would increase transport by AbcA according to our model.

In S. aureus, PSMs have acquired multiple virulence-associated roles, making PSMs important targets in therapeutic strategies to combat certain types of MRSA infections. Our study suggests that blocking PSM efflux by inhibiting PmtCD or AbcA offers distinct advantages. Blocking PmtCD should result in antimicrobial effects from both self-produced PSMs and mammalian antimicrobial peptide accumulation in bacterial cell membranes. In contrast, blocking AbcA may yield an antivirulence approach with reduced antimicrobial effects that result in a weaker selective pressures to develop resistance to AbcA-blockade (40). Furthermore, our study pinpoints AbcA as a target for treatment of atopic dermatitis given the importance of δ-toxin in promoting allergic responses (41) and that AbcA is the dominant δ-toxin transporter.

Materials and Methods

Bacterial Strains and Growth Conditions.

Except where indicated, S. aureus strains (SI Appendix, Table S1) were routinely grown at 37 °C shaking at 180 rpm in tryptic soy broth (TSB) with glucose or statically on tryptic soy agar. Chloramphenicol and tetracycline were added when needed to maintain selection of plasmids at 10 and 12.5 µg/mL, respectively. Plasmids used in this study are listed in SI Appendix, Table S2.

Q5 polymerase, T4 PNK, T4 ligase, and restriction enzymes were purchased from New England Biolabs. Oligonucleotides (SI Appendix, Table S3) were custom synthesized by Sigma Aldrich.

Construction of PSMα Inducible Construct.

pTX hld+psmα1-4 was derived from pTX psmα1-4 plasmid (17). The hld locus was amplified from LAC genomic DNA using primers hld BamHI Forward and hld BamHI Reverse. The product and pTX psmα1-4 were digested with BamHI, ligated, and S. aureus RN4220 cells were transformed with the ligation product. A single colony harboring the construct with hld in the correct orientation to achieve expression was confirmed by Sanger sequencing.

Construction of Transporter Encoding Plasmids.

pRB473 pmtCD was constructed by deleting the pmtAB coding sequence by inverse PCR from pRB473 pmtABCD (17). However, this construct failed to maintain native protein levels of PmtD by western blot (42) in the LAC Δpsmαβhld ΔpmtABCD ΔabcA background. Therefore, we screened several heterologous promoters driving expression of pmtCD for PmtD levels and chose a short segment (P_int) derived from the intergenic region between pbp4 and abcA. Both strands of P_int were synthesized, annealed, and inserted into pRB473 at the EcoRI and KpnI sites. The pmtABCD operon was subcloned from pRB473 pmtABCD downstream of P_int using the KpnI and XbaI restriction sites. The pmtAB coding sequences were deleted by inverse PCR to yield pRB473 P_int-pmtCD. The abcA open reading frame (ORF) and upstream region, as previously reported (18), was cloned into pRB473.

The pmtCD and abcA orthologs from S. epidermidis and S. haemolyticus were amplified using genomic DNA from the corresponding species and subcloned into pRB473.

Construction of Chromosomal Deletions in S. aureus LAC.

Chromosomal deletions were conducted using pIMAY (43). Briefly, Q5 polymerase was used to amplify the targeted ORF plus 1 kb homology arms using the indicated oligos (SI Appendix, Table S3) and LAC genomic DNA as template. The PCR product was inserted blunt into the EcoRV restriction site of pIMAY. The ORF was deleted by inverse PCR using Q5 polymerase and the indicated oligos. All constructs were confirmed by restriction digestion and Sanger sequencing.

The parent strain was transformed with the pIMAY construct by electroporation. Single colonies were passaged once at 28 °C, three times at 37 °C, and once more at 28 °C. Plasmid excision was selected for by growing cells in the presence of 1 µg/mL anhydrotetracycline. Single colonies were screened for loss of chloramphenicol resistance and for the desired deletion by PCR. Deletions were further confirmed by Sanger sequencing of the PCR product.

Cell-Based Secretion Assay.

PSM secretion was assayed as before (19) with modifications. Starter cultures were diluted into TSB without glucose supplemented with chloramphenicol and tetracycline to an OD₆₀₀ = 0.200. Cells were incubated, shaking at 37 °C, 180 rpm for 4 h. D-xylose was added to 0.5% (w/v), and cells were cultured for one additional hour. Cells were pelleted by centrifugation at 13,200 rpm for 2 min in a microcentrifuge, and the supernatants (extracellular fraction) were collected and analyzed for PSMs by injecting 100 µL into a HPLC (Agilent 1260) connected inline to a quadrupole mass spectrometer (MS, Agilent 6120) performed exactly as before (44) using a 2.1 × 5 mm C8 column with increasing gradients of acetonitrile (ACN) + 0.1% trifluoroacetic acid (TFA) to wash the column and elute PSMs. Peptides were detected based on elution time and ion m/z. PSM secretion was quantified by summing the extracted ion chromatogram (EIC) peak area from two ionized (m/z) species per PSM.

PSM Standard Curves.

Synthetic peptides with N-formylation of each α-type PSM (Peptide 2.0) were used to generate standard curves and test for a linear response. Lyophilized peptides were resuspended in DMSO to 10 mg/mL. Concentration series were prepared in 10% ACN + 0.1% TFA and analyzed by HPLC-MS. For each PSM, the EIC peak area as a function of PSM concentration was fit to a linear model ( ${\mathit{EIC}}_i={\beta }_i\ast \left[{\mathit{PSM}}_i\right]$ ), and the slopes ( ${\beta }_i$ ) of the model fits were used to generate figures showing absolute PSM concentration.

PSM Spike-Ins of Culture Supernatants.

Culture supernatants of LAC Δpsmαβhld, ΔpmtABCD, ΔabcA pTX hld-psmα1-4, and either pRB473 P_int-pmtCD or pRB473 abcA were prepared from cultures just before the addition of D-xylose by centrifugation. A cocktail of synthetic S. aureus α-type PSMs was mixed to 20X concentration in 50% ACN + 0.1% TFA and spiked into culture supernatants to 1X at either 50 or 250 nM per PSM (2.5% ACN + 0.005% TFA). PSMs in culture supernatants were measured by HPLC-MS.

Transport Preference Metric.

Individual PSMs have indistinguishable ionization efficiencies regardless of sample origination, and detection by HPLC-MS is linear relative to PSM concentration (Fig. 1 B and C). Thus, we compared the relative secretion of each PSM between cells encoding pmtCD or abcA by calculating the transport probability for each PSM ( $i$ ) through PmtCD as

$$P_i\left(pmtCD\right)=\frac{{\mathit{EIC}}_{i,pmtCD}}{{\mathit{EIC}}_{i,pmtCD}+{\mathit{EIC}}_{i,abcA}}.$$

The use of the EIC is more direct than absolute PSM concentrations, which rely on transformations using standard curves that, moreover, would mathematically cancel out.

Logistic regression is commonly used to regress probabilities on quantitative data. To test the association of PSM transport through PmtCD with PSM hydrophobicity (GRAVY score) and $P_i\left(pmtCD\right)$, we performed a logistic transformation using the logit function:

$$\begin{array}{c}logit\left(P_i\left(pmtCD\right)\right)=\log \left(\frac{P_i\left(pmtCD\right)}{1-P_i\left(pmtCD\right)}\right)={\beta }_0+{\beta }_1\ast {\mathit{GRAVY}}_i\mathit{.}\hfill \end{array}$$

Significance was determined using the F test and the null hypothesis of ${\beta }_1=0$.

LL-37 Sensitivity Assay.

LL-37 was purchased from Bio-Synthesis (lot #P6417-1), resuspended to 1 mg/mL in 0.01% acetic acid, and stored at −20 °C until needed. Starter cultures supplemented with chloramphenicol were diluted to an OD₆₀₀ = 0.1 in specialized bicarbonate-containing media (24, 45) supplemented with chloramphenicol. Cells were cultured shaking for 2.5 h at 37 °C after which cells were collected by centrifugation, resuspended with fresh bicarbonate-containing media without antibiotics to an OD₆₀₀ = 0.1 (10⁸ CFUs/mL), de-clumped by repeated (5×) passage through a 25 G blunt needle, and treated at 37 °C for 3 h with 150 µg/mL LL-37. Viable cells were enumerated by counting single colonies that formed overnight after plating serial dilutions of the treated cells.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.