Membrane selectivity and pore formation of SprA1 and SprA2 hemolytic peptides from Staphylococcus aureus type I toxin–antitoxin systems

Abstract

SprA1 and SprA2 are small hydrophobic peptides that belong to the type I toxin–antitoxin systems expressed by Staphylococcus aureus. Both peptides induce S. aureus death when overexpressed. Although they share 71% of amino acids sequence similarity, SprA2 exhibits stronger hemolytic activity than SprA1. In this study, we investigated the mode of action of these toxins on both prokaryotic‐like and eukaryotic‐like membranes. We first confirmed that SprA2, like SprA1, is an alpha‐helical peptide located at the S. aureus membrane. By overexpressing each toxin, we demonstrated that SprA1 forms stable pores in the S. aureus membrane, evidenced by concomitant membrane depolarization, permeabilization and ATP release leading to growth arrest, whereas SprA2 forms transient pores, causing concomitant membrane depolarization, ATP release, and growth arrest. We showed that the unique cysteine residue present in SprA1 and SprA2 is required for toxicity through disulfide bond formation. Next, we found that both synthetic peptides induce slight leakage in anionic DOPC–DOPG lipid vesicles mimicking prokaryotic membranes, concomitant with lipid vesicles aggregation and/or fusion. Moreover, we observed that SprA1 permeabilizes S. aureus protoplasts, via its ability to form stable pores, whereas SprA2 permeabilizes and lyses them. However, no permeabilization of intact bacteria was detected after the addition of SprA1 and SprA2 in the extracellular medium. Finally, we confirmed that SprA2 has strong activity on zwitterionic DOPC lipid vesicles mimicking eukaryotic membranes, without inducing aggregation. This work highlights the strong selectivity of SprA2 for eukaryotic membranes, suggesting that this toxin may play a role in S. aureus virulence.

When overexpressed in Staphylococcus aureus, the type I toxins SprA1 and SprA2 form membrane pores, with SprA1 creating stable pores and SprA2 forming transient ones. Both induce concomitant membrane depolarization, ATP release, and growth arrest, while only SprA1 causes membrane permeabilization due to its stable pore formation. SprA2 shows a strong selectivity for eukaryotic‐like membranes, consistent with its higher hemolytic activity, suggesting a potential role in S. aureus virulence.

Abbreviations

aTc

anhydrotetracycline

DOPC

1,2‐dioleoyl‐sn‐glycero‐3‐phosphocholine

DOPG

1,2‐dioleoyl‐sn‐glycero‐3‐phospho‐(1′‐rac‐glycerol)

LUVs

Large Unilamellar Vesicles

P/L

peptide/lipid

TA

toxin–antitoxin

Introduction

Toxin–antitoxin (TA) systems are widespread genetic modules in bacterial genomes, typically composed of two genes: one encoding a toxin that causes growth arrest or cell death when overexpressed, and another encoding an antitoxin that inhibits the toxin's activity. These systems are classified into eight types according to the nature of the antitoxin and its mode of action. The antitoxin is a RNA in type I, III, and VIII TA systems and a protein in the others. The antitoxin inhibits toxin mRNA translation in types I, V, and VIII, sequesters the toxin in types II and III, modifies or degrades the toxin in types VI and VII, or competes with the toxin for its target in type IV. Type I systems are characterized by an RNA antitoxin that binds to the toxin mRNA to inhibit its translation or to degrade it. A decrease in the antitoxin/toxin ratio, generally caused by specific stress conditions, leads to toxin excess and toxicity [1]. The mechanisms of action of type I toxins are diverse. For instance, SymE and RalR, located in the cytosol, act as a ribonuclease and an endonuclease, respectively [2, 3]. Other type I toxins are small membrane peptides whose overexpression can induce membrane perturbations, such as membrane depolarization or membrane permeabilization, and/or morphological changes [4]. Although many type I TA systems have been discovered in bacterial genomes, few of them have a known function [5]. Plasmid‐encoded type I TA systems are shown to ensure plasmid maintenance through “plasmid segregational killing” mechanism, but the role of chromosomal type I TA systems is still a matter of debate [1, 6]. Overall, they participate in mobile genetic element maintenance, phage infection, antibiotic persistence, stress adaptation, or biofilm life cycle [2, 7, 8, 9, 10, 11, 12, 13].

In this work, we focused on two type I TA systems, sprA1/SprA1_AS and sprA2/SprA2_AS, expressed by Staphylococcus aureus. S. aureus is an opportunistic mammalian pathogen, which can be multiresistant to antibiotics, and is one of the most widespread pathogens responsible for severe invasive infections such as pulmonary, cardiovascular, or osteoarticular infections [14]. Several TA systems, including type I systems, have been discovered in S. aureus [15]. Six type I TA systems have been described: sprA1/SprA1_AS, sprA2/SprA2_AS, sprG1/SprF1, sprG2/SprF2, sprG3/SprF3, and the tripartite TMCS [16, 17, 18, 19, 20, 21, 22]. To date, no studies have investigated the mechanism of action of these toxins, except for the sprG1/SprF1 type I TA system. Overexpression of the sprG1 toxin gene leads to the production of two toxic peptides, SprG1₃₁ and SprG1₄₄, which cause mesosome formation and membrane depolarization without direct permeabilization [16]. Here, we used these findings as a point of comparison for our two systems of interest, sprA1/SprA1_AS and sprA2/SprA2_AS. The sprA1/SprA1_AS system, present in two to five copies depending on the S. aureus strain, is located on the SaPI3 pathogenicity island [23]. Only one core genome copy, sprA2/SprA2_AS, is expressed in the S. aureus Newman strain. The sprA1 and sprA2 genes share 75% of nucleotide sequence identity in the Newman strain [17]. These TA systems consist of a cis‐encoded antisense antitoxin that represses toxin mRNA translation in trans [17, 22]. No cross‐regulation has been identified between these two type I TA systems [17]. SprA1 and SprA2 toxins are small cationic peptides with a highly hydrophobic N‐terminal part that halt bacterial growth when overexpressed [17, 24]. The SprA1 peptide has an alpha‐helical structure, predicted to be transmembrane [24]. This was confirmed by the determination of its structure via ¹H NMR in an organic solvent (CDCl₃/CD₃OD), used as a model for a hydrophobic environment, as well as by molecular dynamics simulations of its interaction with the membrane [24]. Synthetic SprA1 and SprA2 peptides, previously called PepA1 and PepA2, are hemolytic, with a higher activity for SprA2, but they are not efficiently antimicrobial against Gram‐positive bacteria S. aureus and Gram‐negative bacteria E. coli [17, 24].

The first aim of this work was to determine the mechanisms of action of these toxins by assessing the chronology of membrane effects following SprA1 or SprA2 overexpression. We demonstrated that SprA2 is structured as an alpha helix and located at the S. aureus membrane, like SprA1. We showed that SprA1 forms large pores in the S. aureus membrane, causing concomitant membrane depolarization and permeabilization and ATP release leading to growth arrest, whereas SprA2 forms transient pores in the S. aureus membrane, causing membrane depolarization, ATP release, and growth arrest. Then, we analyzed the membrane interactions of SprA1 and SprA2 by comparing their ability to permeabilize model membranes, protoplasts, or whole cells. We showed that SprA1 and SprA2 lyse anionic and zwitterionic lipid vesicles and protoplasts, but not whole bacteria. As expected, we demonstrated that SprA2 is more active on zwitterionic lipid vesicles mimicking eukaryotic membranes than on anionic lipid vesicles mimicking prokaryotic membranes. This aligns with the previously described higher hemolytic activity of SprA2 than that of SprA1 [17].

Results

SprA2 is a cationic membrane peptide with an alpha‐helical structure

Like most type I toxins, SprG1₃₁, SprG1₄₄, SprA1, and SprA2 are mostly hydrophobic peptides, with SprG1₃₁ having the highest hydrophobicity index and SprA2 the lowest (Table 1). The alpha‐helical structures of SprA1, SprG1₃₁, and SprG1₄₄ were previously resolved by ¹H NMR in model hydrophobic environments (Table 1, Fig. 1A) [4, 24], whereas the structure of SprA2 is unknown. According to the Jpred4 secondary structure predictor, SprA2 is predicted to have an alpha‐helical structure (Table 1, Fig. 1A) [25]. To confirm this prediction, we measured the circular dichroism spectrum of SprA2 in the presence of different percentages of isopropanol, as membrane‐active peptides, like SprA1, are generally unstructured in aqueous solutions [24, 26]. Our results show that SprA2 is poorly structured in water (Fig. 1B). The extent of structured peptide increases with the concentration of isopropanol, reaching a plateau at 30% of isopropanol (Fig. 1B). Given its sequence properties, we expected SprA2 to locate in the membrane, similar to SprA1, SprG1₃₁, and SprG1₄₄ [19, 24]. To investigate the subcellular localization of SprA2, we performed cell fractionation on the S. aureus reference strain HG003 carrying pCN35ΩsprA2‐3xFLAG, where a 3xFLAG sequence was cloned downstream the M1 methionine residue of SprA2 (Table 1). Western blot analysis using anti‐ATPase or anti‐SarA antibodies validated the purity of the membrane or the cytosol fractions, since ATPase was detected in the membrane and SarA in the cytosol (Fig. 1C). Immunoblots, using anti‐FLAG antibodies, confirmed the membrane localization of SprA2 (Fig. 1C). Additionally, SprA2 was detected in the extracellular fraction, as previously shown [17]. These results confirm that SprA2, like SprA1, SprG1₃₁, and SprG1₄₄, is a cationic membrane peptide with an alpha‐helical structure in a membrane environment.

Table 1.

Amino acid sequence, net charge and hydrophobicity of SprA1, SprA2, SprG1₃₁, and SprG1₄₄ peptides. Alpha‐helices were highlighted in yellow. For SprA2, the alpha helix position was predicted with Jpred4 server (https://www.compbio.dundee.ac.uk/jpred/). Hydrophobic amino acids are shown in red, positively charged amino acids in blue, negatively charged amino acids in green and cysteine in yellow. Net charge at pH 7 and hydrophobic index with Kyte Doolittle scale were calculated with R package « Peptides » [57].

Fig. 1.

SprA1, SprA2, and SprG1₃₁ display an alpha‐helical transmembrane domain. (A) Alpha helical representations were created thanks to HeliQuest software (https://heliquest.ipmc.cnrs.fr/index.html) [56] and 3D structure of SprA1 (PBD: 4B19, CDCl₃/CD₃OH) and SprG1₃₁ (PDB: 7NS1, H₂O/isopropanol) solved by ¹H NMR. (B) Circular dichroism spectra of SprA2 at 25 μm in 0 to 50% of isopropanol. (C) Staphylococcus aureus HG003 WT strain carrying pCN35ΩsprA2‐3xFLAG was cultivated in TSB medium until exponential growth phase. After cell fractionation, the expression of SprA2‐3xFLAG was analyzed by western blot. SarA and ATPase expression were monitored as cytosolic (C) and membrane (M) controls, respectively. W and E refers to cell wall and extracellular fractions, respectively. Coomassie blue staining was used as the loading control. The gel is representative of two biological replicates (n = 2).

SprA1 and SprA2 halt S. aureus growth, depolarize its membrane, and promote the release of ATP into the extracellular environment upon overexpression

The mechanism of action of SprG1₃₁ and SprG1₄₄ has already been investigated through overexpression in S. aureus [16]. In this study, we aimed to explore the membrane perturbations induced by SprA1 and SprA2 toxins under overexpression and to compare them with SprG1₄₄ and SprG1₃₁ peptides as controls. The S. aureus HG003 strain was transformed with a pALC plasmid containing either the sprA1 or sprA2 gene under the control of an aTc‐inducible promoter. Northern blots confirmed the overexpression of these genes following aTc induction (Fig. 2A). By measuring OD₆₀₀, we validated that overexpression of sprA1 and sprA2 results in growth arrest, as previously shown in the Newman strain, while no impact in growth was measured without aTc induction [17, 24] (Fig. 2B–D). This growth inhibition occurs 25 min after aTc induction (Fig. 2D). Since SprA1 and SprA2 peptides localize to the S. aureus membrane, we first monitored their ability to induce membrane depolarization at various time points following aTc induction. We used DiBAC₄(3), a dye that enters bacterial cells when the membrane is depolarized and emits fluorescence by binding with hydrophobic intracellular proteins or lipids [27]. Significant membrane depolarization was observed 15 min after aTc induction, while no membrane depolarization was seen in the absence of toxins (Fig. 2E,F). Since previous work on SprG1₄₄ and SprG1₃₁ was performed on the S. aureus N315 strain, which lacks the endogenous sprG1/sprF1 TA system, we confirmed this by inserting sprG1 and sprF1 genes into the HG003 strain using the pALC plasmid (Fig. 2G). After aTc induction, sprG1 expression leads to membrane depolarization within 20 min (Fig. 2H). Nisin also induced membrane depolarization, as previously shown [28] (Fig. 2H).

Fig. 2.

SprA1 and SprA2 overexpression induces membrane depolarization leading to growth inhibition of Staphylococcus aureus. (A) S. aureus HG003 strains carrying pALC, pALCΩsprA1, or pALCΩsprA2 were grown in MH medium until exponential growth phase and incubated in absence (−) or in presence (+) of 0.25 μm aTc. After RNA extraction, northern blot analysis was done on sprA1 and sprA2 expression with 5S rRNA used as the loading control. The gel is representative of one experiment (n = 1), serving only to verify the overexpression of RNAs by the strains. (B–F) S. aureus HG003 strains carrying pALC, pALCΩsprA1, or pALCΩsprA2 were grown in MH medium until exponential growth phase and incubated with or without 0.25 μm aTc. (B) Growth kinetics of S. aureus HG003 strains. (C, D) At each time point after aTc induction or not, OD₆₀₀ were measured. (E, F) At each time point after aTc induction or not, bacteria were incubated with 0.5 μg·mL⁻¹ DiBAC₄(3). DiBAC₄(3) fluorescence values were normalized with OD₆₀₀. (G) S. aureus HG003 strains carrying pALC or pALCΩsprG1/sprF1 were incubated in absence or presence of 0.25 μm aTc. 30 min after induction, RNAs were extracted and northern blot analysis was performed to detect sprG1 or SprF1 RNAs. 5S rRNA was used as a loading control. The gel is representative of one biological replicates (n = 1), serving only to verify the overexpression of RNAs by the strains. (H) S. aureus HG003 strains carrying pALC or pALCΩsprG1/sprF1 were grown in MH medium until exponential growth phase and incubated in absence or presence of 0.25 μm aTc or 12.5 μg·mL⁻¹ nisin. At each time point, bacteria were incubated with 0.5 μg·mL⁻¹ DiBAC₄(3). DiBAC₄(3) fluorescence values were normalized with OD₆₀₀. Error bars show the means and standard deviations of three biological replicates (n = 3). Statistical significance was calculated with the two‐way ANOVA with Tukey's correction. *P < 0.05; ****P < 0.0001. AU, arbitrary unit.

As the ATP synthase needs the proton motive force to function, we measured intracellular ATP levels using a luciferase kit [29]. Results showed a significant decrease in intracellular ATP levels 25 min after aTc induction of SprA1 and SprA2, whereas no decrease was measured without toxins or without aTc induction (Fig. 3A,B). A similar drop in intracellular ATP levels was observed after overexpression of SprG1₃₁ and SprG1₄₄ and after incubation with nisin [16] (Fig. 3C). To evaluate whether this ATP depletion was due to ATP release, we quantified extracellular ATP levels after induction of SprA1 and SprA2. As a positive control, we confirmed that nisin, a pore‐forming toxin, induces significant ATP release [30] (Fig. 3D). While no increase in extracellular ATP levels was detected without toxins or aTc induction, SprA1 and SprA2 overexpression led to significant ATP release after 25 min (Fig. 3E,F). A significant drop in extracellular ATP was observed 60 min after SprA2, SprG1₃₁, and SprG1₄₄ inductions, probably as a consequence of ATP synthesis inhibition (Fig. 3D,F). Conversely, no ATP release was observed 25 min following SprG1₄₄ and SprG1₃₁ overexpression, as previously shown (Fig. 3D) [16]. We also observed that ATP levels in the extracellular fraction are significantly higher following nisin addition compared with SprA1 or SprA2 overexpression, while the increase in DiBAC₄(3) fluorescence and the decrease in intracellular ATP luminescence are of similar magnitudes (Figs 2F,H, 3D,F). The large amount of extracellular ATP observed after the addition of nisin is likely due to the accumulation of ATP in the extracellular medium over time, as nisin rapidly induces permeabilization of the S. aureus membrane [28]. This does not appear to be the case when the SprA1, SprA2, SprG1₄₄, and SprG1₃₁ toxins are overexpressed.

Fig. 3.

SprA1 and SprA2 overexpression causes an ATP leakage concomitant with membrane depolarization and growth inhibition in Staphylococcus aureus. Staphylococcus aureus HG003 strains carrying pALC, pALCΩsprA1, pALCΩsprA2, or pALCΩsprG1/sprF1 were grown in MH medium until exponential growth phase and incubated with or without 0.25 μm aTc or 12.5 μg·mL⁻¹ nisin, used as positive control. (A–C) At each time point after aTc induction or not, bacteria were incubated with BacTiter‐Glo microbial Cell Viability Assay for ATP quantification. Luminescence values were normalized with OD₆₀₀. (D–F) At each time point after aTc induction or not, OD₆₀₀ were measured and bacteria were then centrifuged. The supernatant (extracellular fraction) was then incubated with BacTiter‐Glo microbial Cell Viability Assay for extracellular ATP quantification. Luminescence values were normalized with OD₆₀₀. Error bars show the means and standard deviations of three biological replicates (n = 3). Statistical significance was calculated with the two‐way ANOVA with Tukey's correction. *P < 0.05; **P < 0.01; ****P < 0.0001. AU, arbitrary unit.

Our results demonstrate that the overexpression of SprA1 and SprA2 induces membrane depolarization followed by a depletion of intracellular ATP level due to ATP release. This, in turn, results to growth arrest in S. aureus. Our findings suggest that these two toxins could permeabilize the S. aureus membrane via a pore‐forming mechanism, unlike the SprG1₄₄ and SprG1₃₁ toxins.

SprA1 entirely permeabilizes the S. aureus membrane, unlike SprA2

As mentioned above, the release of ATP outside the bacteria is likely associated with membrane permeabilization. To evaluate membrane integrity, we used the SYTOX Green dye, which enters cells upon membrane permeabilization and emits fluorescence when binding to DNA [31]. Membrane permeabilization kinetics were performed (Fig. 4A). As expected, strong SYTOX Green fluorescence was detected immediately after nisin addition (Fig. 4A,B), consistent with prior findings indicating nisin‐induced membrane permeabilization through stable pore formation [28]. Vancomycin, a non‐pore forming peptide that inhibits cell wall synthesis [32], elicited weaker and delayed membrane permeabilization (Fig. 4A,B). No increase in fluorescence was detected in the absence of toxins, and similar fluorescence profiles were obtained for SprG1₄₄ and SprG1₃₁ upon overexpression, in agreement with previous data (Fig. 4A) [16]. Notably, significant membrane permeabilization occurred 15–20 min after SprA1 induction, aligning with the timing of membrane depolarization observed earlier (Figs 2H, 4B). This suggest that SprA1, like nisin, forms stable pores in the S. aureus membrane when present at sufficient concentrations, consistent with extracellular ATP release (Fig. 3F). Conversely, progressive and minimal membrane permeabilization was measured after SprA2 induction, starting 60 min after aTc induction (Fig. 4B).

Fig. 4.

SprA1 triggers membrane permeabilization concomitant with membrane depolarization in Staphylococcus aureus, unlike SprA2. Staphylococcus aureus HG003 strains carrying pALC, pALCΩsprA1 or pALCΩsprA2 or pALCΩsprG1/sprF1 were grown in MH medium until exponential growth phase. (A, B) Bacteria were incubated with 5 μm SYTOX green probe for 25 min at 37 °C to stabilize fluorescence and incubated with 0.1% ethanol, as vehicle control (A), 0.25 μm aTc (B), 12.5 μg·mL⁻¹ nisin or 28 μg·mL⁻¹ vancomycin, used as positive controls (A, B). (C–F) Bacteria were incubated with 0.1% ethanol, as vehicle control (C–E), 0.25 μm aTc (D–F), 12.5 μg·mL⁻¹ nisin or 28 μg·mL⁻¹ vancomycin (C–F). At each time point, bacteria were incubated 15 min with 5 μm SYTOX Green. After centrifugation, the fluorescence of the bacterial pellet resuspended in MH medium (intracellular fraction) (C, D) and the fluorescence of the supernatant (extracellular fraction) (E, F) were measured. Fluorescence values were normalized with OD₆₀₀. Error bars show the means and standard deviations of three biological replicates (n = 3). Statistical significance was calculated with the two‐way ANOVA with Tukey's correction. ***P < 0.001; ****P < 0.0001. AU, arbitrary unit.

To differentiate between bacterial cell lysis and/or membrane permeabilization contributing to the increase in SYTOX Green fluorescence, we separated intracellular and extracellular fractions and measured SYTOX Green fluorescence in each fraction (Fig. 4C–F). As expected, pore‐forming nisin induced membrane permeabilization without cell lysis, occurring 30 min after incubation. However, an increase in SYTOX Green fluorescence in both fractions was measured 120 min after nisin addition (Fig. 4D,F). This is in agreement with previous study showing that nisin first forms pores into the membrane, and then induces cell autolysis [33]. In contrast, vancomycin did not permeabilize or lyse bacterial cells within 30 min (Fig. 4D,F). After 120 mins of incubation, vancomycin caused a moderate increase in SYTOX Green fluorescence in the intracellular fraction and a significant increase in the extracellular fraction (Fig. 4D,F), in line with the induction of autolysis [34]. No variation in fluorescence was observed in intracellular and extracellular fractions after 30 min and 120 min without toxins (Fig. 4C,E). At 30 min after aTc induction, only SprA1 induced membrane permeabilization without cell lysis, evidenced by an increase in SYTOX Green fluorescence in the intracellular fraction (Fig. 4D,F), in agreement with kinetic observations (Fig. 4B). Two hours (120 min) after aTc induction, SprA1 induced both membrane permeabilization and cell lysis, as indicated by an increase in SYTOX Green fluorescence in both intracellular and extracellular fractions (Fig. 4D,F). Likewise, SprA2 triggered membrane permeabilization and cell leakage 120 min after aTc induction, like SprG1₄₄ and SprG1₃₁, leading to significant increase in SYTOX Green fluorescence in both fractions (Fig. 4D,F).

Altogether, our results demonstrate that SprA1, SprA2, SprG1₃₁, and SprG1₄₄ have distinct mechanisms of action. SprA1 forms stable pores in the S. aureus membrane leading to simultaneous membrane depolarization, permeabilization, and ATP release. SprA2 forms transient pores, triggering membrane depolarization, and ATP release. SprG1₃₁ and SprG1₄₄ do not form pores in the S. aureus membrane, as they induce membrane depolarization without ATP release. These membrane perturbations ultimately lead to S. aureus lysis.

SprA1 and SprA2 require a cysteine residue and disulfide bond formation for optimal toxicity

Some peptides, such as the E. coli toxin HokB, are known to form pores in the membrane when dimerized, notably through disulfide bond [13]. We first investigated whether SprA1 and SprA2 could dimerize thanks to their unique cysteine residue (Table 1). Northern blots were performed to confirm the expression of the sprA1‐3xFLAG and sprA2‐3xFLAG mRNA in HG003 S. aureus strains carrying pCN35ΩsprA1‐3xFLAG and pCN35ΩsprA2‐3xFLAG expressing the flagged toxins (Fig. 5A). We then compared the western profiles of these HG003 S. aureus strains with or without β‐mercaptoethanol treatment, which reduces disulfide bonds. Without β‐mercaptoethanol, both peptides dimerized and even multimerized, whereas treatment with β‐mercaptoethanol resulted in a single band (Fig. 5B). To validate these results, we constructed the plasmids pCN35ΩsprA1‐3xFLAG‐C ₁₅ S and pCN35ΩsprA2‐3xFLAG‐C ₁₆ S, where the cysteine of SprA1 and SprA2 was replaced by serine to prevent disulfide bond formation. Expression of the sprA1‐3xFLAG‐C ₁₅ S and sprA2‐3xFLAG‐C ₁₆ S mRNA was also confirmed by Northern blot (Fig. 5A). A comparison of western blots with or without β‐mercaptoethanol showed no dimerization of the peptides, in line with the crucial role of cysteine in this dimerization/multimerization (Fig. 5C). Furthermore, to assess the impact of peptide dimerization on toxicity, we cloned sprA1‐C ₁₅ S and sprA2‐C ₁₆ S constructs in pALC, allowing aTc‐inducible expression. Northern blot analysis confirmed sprA1‐C ₁₅ S and sprA2‐C ₁₆ S mRNA expression upon aTc induction (Fig. 5D). Then, we showed no difference in growth was observed without induction (Fig. 5E). Conversely, after aTc induction, toxicity induced by SprA1 and SprA2 toxins is significantly reduced when cysteine is substituted with serine (Fig. 5E).

Fig. 5.

SprA1 and SprA2 dimerization and toxicity after cysteine mutation. (A) S. aureus HG003 strains carrying pCN35, pCN35ΩsprA1‐3xFLAG, pCN35ΩsprA1‐3xFLAG‐C ₁₅ S, pCN35ΩsprA2‐3xFLAG and pCN35ΩsprA2‐3xFLAG‐C ₁₆ S, were grown for 3 h at 37 °C before RNA extraction. Northern blot analysis was performed to detect flagged sprA1 or sprA2 RNAs or their mutants. 5S rRNA was used as a loading control. The gel is representative of one experiment (n = 1) serving only to verify the overexpression of RNAs by the strains. (B, C) S. aureus HG003 strains carrying pCN35ΩsprA1‐3xFLAG and pCN35ΩsprA2‐3xFLAG (B), pCN35ΩsprA1‐3xFLAG‐C ₁₅ S and pCN35ΩsprA2‐3xFLAG‐C ₁₆ S (C) were grown in TSB medium until the exponential growth phase. After protein extraction, western blot assays were performed in presence (+) or absence (−) of β‐mercaptoethanol with 1 μg of protein extracts for pCN35ΩsprA2‐3xFLAG, 5 μg of protein extracts for pCN35ΩsprA2‐3xFLAG‐C ₁₆ S, pCN35ΩsprA1‐3xFLAG and pCN35ΩsprA1‐3xFLAG‐C ₁₅ S. The gel is representative of three technical replicates (n = 3). (D, E) S. aureus HG003 strains carrying pALC, pALCΩsprA1, pALCΩsprA2, pALCΩsprA1‐C ₁₅ S or pALCΩsprA2‐C ₁₆ S were grown in MH medium until exponential growth phase and incubated in absence (−) or presence (+) of 0.25 μm aTc. (D) After RNA extraction, northern blot analysis was done on sprA1 and sprA2 expression with 5S rRNA used as the loading control. The gel is representative of one experiment (n = 1), serving only to verify the overexpression of RNAs by the strains. (E) Growth kinetics of S. aureus HG003 strains in absence (left panel) or presence (right panel) of 0.25 μm aTc. Errors bars show the means and standard deviation of three biological replicates (n = 3).

Overall, our findings provide strong evidence that the cysteine residue is essential for the toxicity of SprA1 and SprA2. This is achieved through disulfide bond formation and peptide dimerization/multimerization, leading to pore formation in S. aureus membrane.

SprA1 and SprA2 permeabilize both prokaryotic and eukaryotic mimicking membranes, with SprA2 exhibiting higher activity on eukaryotic mimicking membranes

We previously reported that synthetic peptides SprA1 and SprA2 exert cytotoxic effects on human erythrocytes but lack antimicrobial activity [17]. However, in this study, we found that these peptides can form transient or stable pores in S. aureus membrane when overexpressed. To better understand these peptide–membrane interactions, we investigated the ability of synthetic peptides to interact with and permeabilize membrane models by varying Peptide/Lipids (P/L) ratios. We used large unilamellar vesicles (LUVs) around 100 nm in diameter, either made of DOPC‐DOPG (3:1) to mimic the anionic prokaryotic cell membrane, or composed of DOPC alone, mimicking the zwitterionic eukaryotic cell membrane. For our experiments, we employed calcein leakage assay using lipid vesicles containing calcein dye at a self‐quenching concentration. Permeabilization of these lipid vesicles leads to dye leakage and an increase in fluorescence. We showed that SprA1 and SprA2 are able to permeabilize DOPC‐DOPG (3:1) anionic LUVs and DOPC zwitterionic lipid vesicles, albeit with varying intensity (Fig. 6A). While SprA1 exhibited similar low permeabilization efficiency toward zwitterionic and anionic lipid vesicles, SprA2 showed significantly higher efficiency on zwitterionic membranes. Nisin, whose its main target is the cell wall precursor Lipid II [35, 36], exhibits low activity against lipid vesicles compared with SprA1 and SprA2 (Fig. 6A). However, we observed that nisin induces a higher permeabilization of DOPC‐DOPG (3:1) lipid vesicles than for DOPC (Fig. 6A), which is consistent with previous study [37].

Fig. 6.

SprA1 and SprA2 permeabilize both zwitterionic and anionic lipid vesicles, with SprA2 exhibiting higher activity on zwitterionic lipid vesicles. Calcein leakage assays from zwitterionic DOPC or anionic DOPC–DOPG (3:1) LUVs after SprA1 or SprA2 addition. LUVs of zwitterionic DOPC or anionic DOPC–DOPG (3:1), loaded with calcein at self‐quenching concentration, were incubated with SprA1 or SprA2 at different Peptide/Lipids (P/L) molar ratios for 15 min at 22 °C. To lyse all the lipid vesicles, Triton X‐100 was added. (A) The ratio of fluorescence after peptides addition to fluorescence after Triton X‐100 addition is plotted against different P/L molar ratios. Error bars show the means and standard deviations of three independent experiments (n = 3). (B, C) The fluorescence kinetics of one of the experiments are represented over the time.

Examining the kinetics of calcein leakage from DOPC–DOPG (3:1) lipid vesicles revealed an initial burst of leakage immediately after peptide addition, followed by a plateau (Fig. 6B,C). This suggests a “transient permeabilization model,” where pore formation initially releases dye from some vesicles while others remain intact. Possible mechanisms include dissipation of trans‐bilayer asymmetry due to peptides accumulation in one leaflet until they translocate and distribute equally, or vesicle fusion and/or aggregation [38]. Interestingly, the kinetics of calcein leakage from DOPC lipid vesicles showed a burst followed by a gradual increase in calcein fluorescence (Fig. 6B,C), similar to the “hybrid permeabilization model” commonly observed with membrane‐active peptides [38].

Based on these results, we can conclude that the selectivity of SprA2 action appears to depend on the charge of the lipid vesicles, whereas membrane charge does not impact SprA1 activity.

SprA1 and SprA2 aggregate with anionic lipid vesicles, but not with zwitterionic lipid vesicles

As cationic peptides often induce aggregation and/or fusion of anionic lipid vesicles, we measured the turbidity of lipid vesicles solutions after peptide addition (Fig. 7A,B). No change was observed with DOPC zwitterionic lipid vesicles, whereas an increased turbidity was detected for DOPC–DOPG (3:1) anionic lipid vesicles with higher P/L ratios (Fig. 7A,B). Dynamic Light Scattering (DLS) confirmed this aggregation and/or fusion (Fig. 7C,D). This phenomenon arises from the binding of cationic peptides to negatively charged lipid vesicles until neutrality, destabilizing the emulsion and causing an aggregation and/or fusion due to the cancelation of repulsive interactions between the LUVs. Over a period of 1 h, we observed a decrease in turbidity at high P/L ratio for DOPC–DOPG lipid vesicles (Fig. 7E). This decrease correlates with the accumulation of aggregates observed in wells after 1 h (Fig. 7F). Notably, a slight increase in lipid vesicles size (from 100 to 200 nm) was measured by DLS when SprA1 is added at P/L = 0.01 (Fig. 7C). This change in lipid vesicles size caused by SprA1 could be explained by its higher hydrophobicity leading to weak vesicles fusion [39].

Fig. 7.

SprA1 and SprA2 aggregate with anionic lipid vesicles, but not with zwitterionic lipid vesicles. (A, B) Turbidity measurements of solutions containing zwitterionic DOPC (A) or anionic DOPC–DOPG (3:1) (B) LUVs after addition of SprA1 or SprA2 at different Peptide/Lipids (P/L) molar ratios. Each point corresponds to OD₄₃₆ subtracted with the initial OD₄₃₆. Error bars show the means and standard deviations of three measurements. (C, D) Hydrodynamic Diameter D_H (nm) of LUVs of zwitterionic DOPC (C) or anionic DOPC‐DOPG (3:1) (D) after addition of SprA1 or SprA2 at different Peptide/Lipids (P/L) molar ratios, measured by DLS. Error bars show the means and standard deviations of four measurements. (E) OD₄₃₆ were measured for 1 h after incubation of anionic DOPC‐DOPG (3:1) lipid vesicles with SprA1 or SprA2 peptides at different Peptides/Lipids (P/L) molar ratios: P/L = 0.02; P/L = 0.04; P/L = 0.075; P/L = 0.1. (F) Pictures of wells taken 1 h after addition of SprA1 or SprA2 to DOPC–DOPG (3:1) lipid vesicles.

To further investigate this aggregation process, we titrated peptide on anionic DOPC–DOPG (3:1) or zwitterionic DOPC lipid vesicles by ¹H NMR. This experiment could only be carried out for SprA2, which is the only one that can be solubilized in water without cosolvent (Fig. 8). The addition of SprA2 to a 1 mm solution of DOPC–DOPG lipid vesicles failed to detect NMR signals from aromatic protons of the peptide, where such signals are usually easily identifiable (Fig. 8). This absence of peptide NMR signal corresponds to a drastic decrease in peptide mobility. Indeed, this decrease is responsible for a broadening beyond the detection of proton resonances [40, 41, 42]. It should also be noticed that concomitantly, a decrease in the lipid vesicles NMR signals is clearly observed, looking at the peaks corresponding to choline methyl protons and fatty acid terminal methyl protons, suggesting a coaggregation of lipid vesicles and peptides. In contrast, no significant changes were detected in the ¹H NMR spectra of DOPC LUVs after the addition of SprA2 (Fig. 8), consistent with the absence of lipid vesicles aggregation. Regarding peptide NMR signals, as expected, their intensities increase with peptide concentration in solution. Nevertheless, these intensities remain lower than those of free peptide at the same concentration. This difference can be associated with a greater line broadening induced by certain dynamic interactions with LUVs.

Fig. 8.

SprA2 peptide coprecipitates with anionic lipid vesicles, but not with zwitterionic lipid vesicles. (A) Reference ¹H NMR spectra of 150 μm SprA2, 1 mm LUVs of zwitterionic DOPC or anionic DOPC–DOPG (3:1). (B–D) ¹H NMR spectra of 50 μm (B), 100 μm (C) or 150 μm (D) of SprA2 in absence or presence of 1 mm LUVs of DOPC or DOPC–DOPG (3:1). Signals of H₂O, Tris, TSP (trimethylsilylpropanoic acid) are shown in (A). Signals of lipids and SprA2 are labeled on the spectra.

Overall, our results demonstrate that SprA1 and SprA2 effectively permeabilize both anionic or zwitterionic lipid vesicles and induce strong aggregation of anionic lipid vesicles. Our results also highlight that SprA2 exhibits enhanced activity on zwitterionic lipid vesicles, showing the importance of charges for its mechanism of action.

SprA1 permeabilizes bacterial protoplasts, while SprA2 lyses them

We showed that SprA1 and SprA2 exhibit low permeabilizing activity against anionic lipid vesicles, potentially due to their aggregation with these lipid vesicles. This modest activity is in agreement with their lack of antimicrobial effects. However, when overexpressed, these peptides efficiently permeabilize bacterial membranes. To model conditions closer to bacteria than lipid vesicles, we assessed whether SprA1 and SprA2 peptides affect the membrane integrity of S. aureus protoplasts. We evaluated their ability to permeabilize or lyse protoplasts from the S. aureus HG003 strain by measuring, respectively, SYTOX Green fluorescence or OD₆₀₀ (Fig. 9A,B). For comparison, we conducted the same experiment using whole S. aureus cells (Fig. 9C,D). We showed that both peptides permeabilize protoplasts, as evidenced by increased SYTOX Green fluorescence, but not whole bacteria, unlike nisin (Fig. 9A–D). SprA2 induces a strong permeabilization of protoplasts, whereas SprA1 induces a lower permeabilization starting at 10 μm peptide. These effects appear to be concentration‐dependent, with SprA1 showing no effect at 5 μm and SprA2 having maximal effect at 20 μm (Fig. 9A). Moreover, SprA2‐induced permeabilization correlates with protoplasts lysis, evidenced by a decrease in OD₆₀₀, unlike SprA1 (Fig. 9B).

Fig. 9.

SprA1 and SprA2 are able to permeabilize Staphylococcus aureus protoplasts, but are unable to permeabilize whole bacteria. (A, B) Protoplasts of HG003 S. aureus strain were incubated with 5 μm SYTOX green, and then with SprA1 or SprA2. SYTOX Green fluorescence (A) and OD₆₀₀ (B) were measured. To lyse all the protoplasts, 0.5% Triton X‐100 was added. (C, D) S. aureus HG003 WT strain was grown in TSB medium for 3 h at 37 °C and then stained with 5 μm SYTOX Green. After 10 min at 37 °C, bacteria were incubated with 20 μm SprA1, 20 μm SprA2 or 12.5 μg·mL⁻¹ nisin. SYTOX Green fluorescence (C) and OD₆₀₀ (D) were measured every 2 min for 30 min. Error bars show the means and standard deviations of three biological replicates (n = 3).

Altogether, our results demonstrate that SprA1 can permeabilize the S. aureus membrane by forming large pores, but does not trigger lysis of S. aureus protoplasts. In contrast, SprA2 induces lysis of S. aureus protoplasts, likely due to its ability to form small or transient pores that disrupt the S. aureus membrane.

Discussion

With the exception of SymE and RalR, all known type I toxins are small peptides, mostly cationic and hydrophobic. They are predicted to adopt an alpha‐helical structure, although experimental confirmation through circular dichroism or ¹H NMR has been achieved only for a few [4]. Here, we contribute to a new structural description of SprA2, a type I toxin of S. aureus, using circular dichroism. The peptide is mainly unstructured in solution, while its helical content increases with the apolar character of the solvent, indicating that the peptide exhibits the intrinsic propensity to form helical structures in a membrane environment. Despite its relatively low overall hydrophobic index (Table 1), SprA2 is an amphipathic peptide due to the presence of two distinct parts in the peptide: a long, highly hydrophobic N‐terminal helical segment and a relatively short, hydrophilic C‐terminal segment.

In the first part of our study, we examined the interactions of SprA1 and SprA2 with the S. aureus membrane when overexpressed by the bacteria (Fig. 10). Overexpression of these toxins resulted in membrane depolarization followed by a reduction in intracellular ATP levels 15–25 min after induction, coinciding with growth inhibition. Similar results were previously obtained for SprG1₄₄ and SprG1₃₁ toxins [16]. For these four peptides, growth inhibition appears to be due to membrane depolarization. This could be a common mechanism with that of DinQ, IbsC, ShoB, and TisB type I toxins, where initial membrane depolarization and/or depletion of intracellular ATP precede growth inhibition, considering the experimental time required to measure membrane depolarization [43, 44, 45]. However, the interaction of peptides with the bacterial membrane leading to membrane depolarization appears to differ between SprA1, SprA2, SprG1₄₄, and SprG1₃₁ toxins. Indeed, we observed significant membrane permeabilization at the same timescale as membrane depolarization for SprA1. This suggests that SprA1 forms stable pores in the bacterial membrane, allowing intracellular DNA staining with the SYTOX Green dye. Furthermore, we confirmed that the DNA detected by SYTOX Green is located in the intracellular fraction and results from membrane permeabilization rather than cell lysis, like the pore‐forming peptide nisin. In contrast, SprA2 does not form stable pores on the same timescale as membrane depolarization but induces membrane permeabilization and cell lysis later. The same behavior was observed for SprG1₄₄ and SprG1₃₁ toxins [16]. This delayed cell lysis could be attributed to the induction of autolytic enzymes, as previously shown for several antimicrobial peptides, as nisin and vancomycin, and the type I toxin BsrG [33, 34, 46]. However, by measuring ATP levels in the extracellular medium after membrane depolarization, we demonstrated that an increase in extracellular ATP was detected for SprA1 and SprA2, unlike SprG1₄₄ and SprG1₃₁. The release of ATP by SprA1 confirms its ability to form stable pores. For SprA2, our results suggest that it can form transient pores in the membrane, facilitating ATP release. This is in agreement with a previous study, showing that the antimicrobial peptide Mel4 induces ATP release in small amounts without permeabilizing the membrane to SYTOX Green [47]. The excretion of ATP in small amounts may require transient permeabilization, contrasting with the entry of SYTOX Green molecules into bacteria. In contrast, we previously showed that overexpression of SprG1₄₄ and SprG1₃₁ does not lead to ATP excretion, suggesting that these peptides may form small ionic channels or employ a more “detergent‐like” mechanism [48]. Here, we also showed that, thanks to a mutagenesis study, SprA1 and SprA2 form disulfide bonds crucial for their toxicity. Similar findings were previously reported for the toxin HokB, which also forms pores in the bacterial membrane leading to ATP release [13]. Taken together, our results emphasize the importance of investigating the chronology of membrane perturbations induced by toxin overexpression, particularly through concomitant measurements of membrane depolarization and permeabilization. This approach is essential for initial investigations to decipher the mechanism of action of type I toxins.

Fig. 10.

Timeline of toxic effects of SprA1 (A) and SprA2 (B) under overexpression. Overexpression of SprA1 peptides results in dimerization, followed by membrane depolarization, ATP excretion, and inhibition of ATP synthesis. These effects are accompanied by membrane permeabilization, occurring approximately 15–20 min after aTc induction. Subsequently, cell lysis occurs around 90–120 min post induction. Overexpression of SprA2 peptides also leads to dimerization, followed by membrane depolarization, ATP excretion, and inhibition of ATP synthesis approximately 15–20 min after aTc induction. Membrane permeabilization and cell lysis then occur around 90–120 min post induction.

In the second part of this study, we aimed to validate our data from overexpression conditions using model membranes to evaluate the pore‐forming ability of SprA1 and SprA2. Given the strong hemolytic activity of SprA2, we specifically compared its effects on anionic DOPC‐DOPG (3:1) model membranes, mimicking bacterial membranes, and zwitterionic DOPC model membranes, mimicking mammalian membranes. LUVs were employed as model membranes, as they are largely used to study peptide–membrane interactions, mostly for antimicrobial peptides. We chose this basic model to assess the impact of charge, where lipid composition primary differs in terms of charge density. First, we assessed the ability of SprA1 and SprA2 to permeabilize both types of LUVs using calcein leakage assays. We showed that both types of lipid vesicles exhibited low leakage rates with all peptides except SprA2, which showed significantly higher activity against zwitterionic lipid vesicles (Fig. 11). This correlates with the strong hemolytic activity of SprA2 and highlights the importance of membrane charge in its mechanism of action. This specificity toward eukaryotic membranes may initially seem surprising, given that its positive charges typically suggest better affinity and partitioning with anionic membranes (Table 1). However, several studies have already demonstrated examples of cationic peptides that preferentially permeabilize neutral membranes over anionic ones. For example, the well‐known membrane‐active peptide melittin, from bee venom, is significantly more active on DOPC vesicles than on DOPG ones [49]. This selectivity is due to entirely different mechanisms for these membrane types: While melittin interacts more strongly with DOPG lipids, the peptides accumulate on the surface rather than insert into the membrane and form pores, as they do with DOPC membranes. Consequently, pore formation and subsequent permeabilization occur in DOPC membranes but not in DOPG membranes [49]. Therefore, it is reasonable to propose that SprA2 may function similarly to melittin, which could explain its selectivity for the zwitterionic membrane. We also assessed changes in lipid vesicles turbidity and size after peptides addition. SprA1 or SprA2 induced a high aggregation of anionic DOPC–DOPG (3:1) LUVs, reflecting a coprecipitation of lipid vesicles and peptides. This artifact is common when studying interactions between lipid vesicles and cationic peptides, where peptide binding neutralizes vesicles charges, leading to a colloidal destabilization [38]. Given the low leakage rates, even at high peptide concentrations, and the substantial aggregation, it is reasonable to assume that leakage from anionic lipid vesicles is also influenced by vesicle aggregation. These data indicate that the anionic lipid vesicles model is susceptible to aggregation artifacts. Moreover, both peptides showed no significant activity against anionic membranes, in line with their lack of antimicrobial activity [17]. In contrast, interaction with zwitterionic DOPC LUVs had minimal impact on lipid vesicles size compared with anionic DOPC–DOPG LUVs. SprA2 did not alter the size of DOPC LUVs, whereas SprA1 provoked an increase of 100 nm, approximately the size of these lipid vesicles. This increase could be attributed to SprA1‐induced vesicle fusion and may contribute to membrane permeabilization in this membrane type.

Fig. 11.

Toxic effects of synthetic peptides SprA1 (A) and SprA2 (B) on membranes. Synthetic SprA2 peptides provoke slight permeabilization and aggregation of anionic DOPC–DOPG (3:1) LUVs but induce significant permeabilization of zwitterionic DOPC LUVs. They cause membrane permeabilization and lysis of protoplasts of Staphylococcus aureus but no permeabilization of whole S. aureus cells (right panel). MIC and HC₅₀ values comes from published data in Germain‐Amiot et al. (2019) [17].

Since SprA1 and SprA2 were able to permeabilize bacterial membranes when overexpressed, we used protoplasts to compare these results with lipid vesicles leakage assays. We showed that SprA1 and SprA2 could permeabilize S. aureus protoplasts but not whole bacteria (Fig. 11). This discrepancy could be explained by (a) the presence of the cell wall, which reduces the concentration of peptides reaching and binding to the bacterial membrane, or (b) potential peptide‐induced aggregation/fusion of protoplasts, a phenomenon that could lead to their permeabilization. Moreover, a small fraction of the total cell population may have been permeabilized but remained undetected in our assay. To note, the protoplast assay was performed with approximately 10⁸ cells·mL⁻¹, estimated from the measured OD₆₀₀. Given the approximately 1 μm diameter of spherical S. aureus cells and an estimated surface area per lipid molecule of 0.7 nm², with about 50% of the surface containing proteins, each cell contains approximately 2.2 × 10⁶ lipids, leading to an estimated lipid concentration of about 0.8 μm. Consequently, protoplasts permeabilization occurred at about P/L = 10 for SprA1 and lower for SprA2. These ratios are much higher than those used in the calcein leakage experiments with lipid vesicles, and the different membrane compositions preclude direct comparison of permeabilization ratios between experiments. Additionally, the context is quite different, as bacteria can repair some membrane damage [50]. However, both findings indicate that SprA2 exhibits higher permeabilization activity than SprA1. We also showed that, compared with SprA1, SprA2 induces a strong protoplast permeabilization mainly associated with protoplast lysis. This may seem surprising, as SprA1 was able to entirely permeabilize bacterial cells when overexpressed, unlike SprA2. This difference in permeabilization appears to be due to a difference in mechanism rather than in the strength of activity, as the level of depolarization was similar. For instance, SprG1₃₁ exhibits antibacterial activity when added to the extracellular medium, whereas it does not directly permeabilize the membrane when overexpressed [16, 19]. Moreover, since we do not know whether the peptide concentration at the membrane is the same for SprA1 and SprA2 when overexpressed, making a direct comparison is challenging. Finally, the aggregation state of SprA1 in the medium and its organization in the membrane when added to the extracellular medium may differ from when it is overexpressed, due to its high hydrophobicity.

To conclude, this study provides insights into the mechanism of action of SprA1 and SprA2 upon overexpression, suggesting that SprA1 forms large, stable pores in the S. aureus membrane, whereas SprA2 induces a transient permeabilization. In the literature, two types of pores are commonly cited: the barrel‐stave pore, where the hydrophobic regions of peptides align with hydrophobic core of lipids, and the toroidal‐pore model, where peptides induce a lipid curvature [51]. However, pores formed by peptides are highly variable—they can be stable or unstable, well‐defined or not, and can vary in size and structure [52, 53]. Furthermore, membrane‐active peptides can exhibit different mechanisms and interactions depending on the P/L ratio, lipid type, membrane type, peptide aggregation state, and experimental conditions such as pH and temperature [54]. Further biophysical studies, such as conductance measurements on planar bilayers, are needed to decipher the precise mechanism of action of SprA1 and SprA2. Moreover, the exact peptide concentrations required to induce membrane permeabilization under stress conditions remains to be elucidated. Finally, to better understand the intriguing selectivity of SprA2, future biophysical and dynamic simulation studies should examine its affinity for both types of membrane and determine whether the peptide employs the same mechanism of action depending on lipid charge (anionic or zwitterionic membranes) and lipid polymorphism.

The biological roles of SprA1 and SprA2 peptides remain unknown. The sprA1 gene is located on the SaPI3 pathogenicity island in the S. aureus Newman strain and is overexpressed under oxidative stress [24]. As it can form stable pores in the membrane, SprA1 may be involved in S. aureus antibiotic persistence, as this is the case for the pore‐forming toxins HokB or TisB [8, 13]. The sprA2 gene is located on the core genome and overexpressed under nutritive stress [17]. Given its high hemolytic and cytolytic activities and its excretion in the extracellular medium, this toxin may also contribute to S. aureus virulence. Future studies will assess the biological activity of these peptides to confirm data obtained with synthetic peptides. Moreover, further in vivo studies are warranted to explore the roles of SprA1 and SprA2 in S. aureus virulence and/or antibiotic persistence.

Materials and methods

Strains and plasmid constructions, growth and toxin induction

Bacterial strains, plasmids and primers used in this work are listed in Tables S1–S3. To generate an anhydrotetracycline (aTc)‐inducible construct for sprG1 ₃₁₂ expressing the two sprG1 ₃₁₂ ‐encoded peptides (SprG1₃₁ and SprG1₄₄), a DNA fragment containing the sprG1 ₃₁₂ sequence beginning at the transcription start site and ending at its 3′‐end while including the sprF1 sequence was amplified from HG003 genomic DNA by PCR using KOD Hot Start polymerase (Novagen, Millipore, Molsheim, France). To generate a pCN35 plasmid containing sprA1‐3xFLAG, a DNA fragment containing the sprA1 sequence beginning at the transcription start site and ending at its 3′‐end was amplified from HG003 genomic DNA by PCR using KOD Hot Start polymerase (Novagen, Millipore, Molsheim, France). The 3xFLAG sequence was added downstream the ATG initiation codon. For the “sprA1‐3xFLAG‐C ₁₅ S” mutant, the cysteine residue at the position 15 was replaced by a serine in the plasmid pCN35ΩsprA1‐3xFLAG. For the “sprA2‐3xFLAG‐C ₁₆ S” mutant, the cysteine residue at the position 16 was replaced by a serine in the plasmid pCN35ΩsprA2‐3xFLAG. S. aureus HG003 strains carrying pCN35 plasmid with sprA1‐3XFLAG, sprA2‐3xFLAG, sprA1‐3xFLAG‐C ₁₅ S or sprA2‐3xFLAG‐C ₁₆ S, were grown overnight in tryptic soy broth (TSB, Oxoid, ThermoFisher Scientific, Courtaboeuf, France) with 10 μg·mL⁻¹ chloramphenicol at 37 °C under agitation. They were then diluted at 1/100 in TSB medium with 5 μg·mL⁻¹ chloramphenicol and grown for 4 h at 37 °C under agitation before RNA and protein extractions. S. aureus HG003 strains carrying pALC plasmids were grown overnight in Mueller Hinton broth (MH, Oxoid, ThermoFisher Scientific, Courtaboeuf, France) with 10 μg·mL⁻¹ chloramphenicol. The cultures were then diluted to an optical density at 600 nm (OD₆₀₀) of 0.05 in MH and grown for 2.5 h until the exponential growth phase. The overexpression of SprA1, SprA2 or SprG1₄₄ and SprG1₃₁ was then induced by adding 0.25 μm of aTc initially diluted in 100% ethanol. An equivalent concentration of ethanol was added for the control. For protoplasts preparation, S. aureus HG003 wild‐type was grown overnight in TSB medium at 37 °C under agitation, then diluted at 1/100 in TSB medium and grown for 3 h at 37 °C under agitation.

Protein extractions, cell fractionation, and western blots

After 4 h of culture in TSB medium (OD₆₀₀ = 4), S. aureus HG003 strains carrying pCN35 or pCN35ΩsprA1‐3xFLAG, pCN35ΩsprA1‐3xFLAG‐C ₁₅ S or pCN35ΩsprA2‐3xFLAG or pCN35ΩsprA2‐3xFLAG‐C ₁₆ S were centrifuged 10 min at 15 700 g at 4 °C. The supernatant containing extracellular proteins was incubated with TCA at the final concentration of 10% at 4 °C overnight. The pellet was resuspended in lysis buffer composed of 50 mm Tris–HCl, pH 7.5, 20 mm MgCl₂, supplemented with 37% sucrose, PMSF at 1 mm and lysostaphin (0.1 mg·mL⁻¹). The suspension was then incubated 15 min at room temperature and centrifuged at 1500 g at 4 °C for 8 min. The supernatant containing cell wall proteins was incubated with TCA at the final concentration of 10% at 4 °C overnight. The pellet was dissolved into 50 mm Tris–HCl, pH 7.5, 20 mm MgCl₂ with protease inhibitors (Roche Diagnostics, Merck, Saint‐Quentin Fallavier, France) added to a 1x final concentration and mechanically broken through beads beating using a FastPrep‐24 5G instrument (MP Biomedicals, Illkirch‐Graffenstaden, France). After 1‐min centrifugation at 1500 g and 4 °C, the supernatant was collected and centrifugated 60 min at 46 000 g and 4 °C. The supernatant, containing the cytoplasmic protein fraction, was incubated with TCA at the final concentration of 10% at 4 °C overnight. The pellet, containing the membrane protein fraction, was dissolved in 70 μL of buffer containing 50 mm Tris–HCl, pH 7.5, 20 mm MgCl₂, 0.01% Triton‐X100, and 1× protease inhibitors. After 12 h of incubation at 4 °C for protein precipitation, the extracellular, cell wall, and cytoplasmic fractions were resuspended in 30 μL of buffer loading containing 60 mm Tris HCl pH 6.8, 10% glycerol and 2% SDS. For the western blots, 5 μL of extracellular, cell wall, and cytoplasmic fractions and 0.5 μL of membrane fraction (about 1 μg of proteins for membrane fraction) were mixed with loading buffer containing or not 10% of 14.1 m β‐mercaptoethanol, heated 10 min at 95 °C and loaded onto a 4–16% Tricine SDS/PAGE. Electrophoresis was done at 50 V for 30 min and then at 120 V for 90 min. Proteins were electrotransferred onto Amersham Hybond‐P PVDF membranes (GE Healthcare, Merck, Saint‐Quentin Fallavier, France) at 30 V, 4 °C for 3 h. After blocking, membranes were incubated with monoclonal mouse anti‐FLAG M2 antibody (1:2000; Merck, Saint‐Quentin Fallavier, France), rabbit polyclonal anti‐Micrococcus luteus ATPase (1:1500), [55] or rabbit anti‐S. aureus SarA (1:10 000, produced in our laboratory). Incubation was carried out for 1 h at room temperature (anti‐FLAG) or overnight at 4° C (anti‐ATPase and anti‐SarA), and the membranes were washed and incubated with HRP‐conjugated secondary antibodies for 1 h at room temperature. The membranes were then washed, revealed with an Amersham ECL Plus Western Blotting Detection kit, and scanned with an ImageQuant LAS 4000 imager (GE Healthcare, Merck, Saint‐Quentin Fallavier, France). Coomassie blue staining was done as a loading control.

RNA extractions and northern blot assays

As previously described [16], bacterial cultures were collected 30 min after aTc induction by centrifugation at 4500 g for 10 min at 4 °C. The resulting pellets were promptly frozen at −80 °C. To extract RNA, the pellets were resuspended in 500 μL of a lysis buffer containing 0.5% SDS, 20 mm sodium acetate, and 1 mm EDTA (pH 5.5). Cells were lysed mechanically using bead beating in acidic phenol (pH 4) with a FastPrep‐24 5G system (MP Biomedicals, Illkirch‐Graffenstaden, France). The lysates were centrifuged for 5 min at 15 700 g at 4 °C, and the aqueous layer was recovered. This phase was re‐extracted with an equal volume of acidic phenol and centrifuged again under the same conditions. Subsequently, the aqueous phase was mixed with a 24:1 chloroform/isoamyl alcohol solution, followed by another 5‐min centrifugation at 15 700 g and 4 °C. RNA was precipitated by adding 2.5 volumes of ethanol and 0.1 volume of 3 m sodium acetate (pH 5.2) and incubating at −20 °C. For northern blot analysis, 10 μg of RNA was separated on an 8% urea‐PAGE gel and transferred to a ZetaProbe GT membrane (Bio‐Rad, Marnes‐la‐Coquette, France) using 0.5× TBE buffer (90 mm Tris, 90 mm boric acid, 2 mm EDTA) at 25 V for 2 h. RNA was cross‐linked to the membrane through UV irradiation. Specific RNA probes (10 pmol; sequences detailed in Table S3) were labeled with [γ³²P]‐ATP (5 μCi) using T4 polynucleotide kinase. Membranes were hybridized overnight in ExpressHyb solution (Ozyme, St Cyr l'Ecole, France). Washing steps included two rinses in 2× SSC (0.3 m sodium chloride, 0.03 m sodium citrate, pH 7.0) with 0.05% SDS for 10 min each, followed by two rinses in 0.1× SSC with 0.1% SDS for 10 min. Membranes were then exposed and visualized using a PhosphorImager.

Measurement of S. aureus membrane depolarization and permeabilization

Membrane depolarization assays were performed as previously reported with some modifications [16]. After aTc induction, 2 mL of S. aureus cultures was centrifuged at 6000  g for 2 min at each time point. Bacteria were resuspended in PBS and stained with 0.5 μg·mL⁻¹ DiBAC₄(3) (ThermoFisher Scientific, Courtaboeuf, France) by incubation at 37 °C for 4 min in the dark. Then, OD₆₀₀ and fluorescence were measured using the Synergy 2 microplate reader (Agilent, Les Ulis, France). Fluorescence values were normalized with OD₆₀₀. Membrane permeabilization assays in kinetics were performed in microplates as previously described [16]. For time point measurements of SYTOX Green fluorescence (ThermoFisher Scientific, Courtaboeuf, France) in intracellular and extracellular fractions, bacterial cultures were induced with 0.25 μm aTc or 0.1% ethanol, as vehicle control, or 12.5 μg·mL⁻¹ nisin or 28 μg·mL⁻¹ vancomycin. At each time point, 225 μL of culture were incubated with 25 μL of SYTOX Green at 25 μm for 15 min at 37 °C under agitation. Then, the cultures were centrifuged at 8000  g for 3 min and 100 μL of the supernatant was loaded in a black microplate for fluorescence and absorbance measurements: It is the extracellular fraction. The remaining supernatant is discarded; the pellet is resuspended in 250 μL of MH medium and 100 μL were loaded into a microplate: It is the intracellular fraction. Fluorescence (λ _exc = 488/λ _em = 525) and OD₆₀₀ of each fraction were then measured using the Synergy 2 microplate reader. For all of these time‐point measurements, the indicated “time after induction” is the time of measurement.

Measurement of intracellular and extracellular ATP

Measurements of intracellular and extracellular ATP were performed as previously described [16].

Synthetic peptides

SprA2 peptide was synthetized by Proteogenix (Schiltigheim, France), and analyzed by HPLC and MS. SprA1 peptide was synthetized as described previously (Germain‐Amiot et al., 2019). Both peptides are resuspended in 20% or 25% of isopropanol at 2 mm and stored at −80 °C in protein Low‐bind tubes.

Circular dichroism (CD)

Circular dichroism spectra were obtained using a Jasco J‐815 instrument (Easton, USA) associated with a Peltier device for temperature control, available on the CDTP platform (UAR Biosit 3480 CNRS – US18 Inserm, Univ Rennes, France). Measurements were performed in a continuous scanning mode from 190 to 250 nm with intervals of 0.5 nm, a digital integration time of 1 s with a 0.5 nm bandwidth and a scan speed of 500 nm per min at 20 °C with a quartz cell with a path length of 2 mm. Each spectrum is the mean of 3 successive scans. The CD signals were calculated in units of mean residue molar ellipticity θ (deg·cm²·dmol⁻¹) from millidegrees (mdeg) with the following equation.

$$\theta =\frac{{\theta }_{\mathrm{obs}}}{\mathrm{l}\times \mathrm{C}\times 10\times n}$$

where θ _obs is the CD signal in mdeg, C is the concentration in mol/L, l the pathlength in cm and n the number of residues in the peptide.

Lipid vesicles preparation

Lipid vesicles were prepared by solvent evaporation followed by an extrusion. Lipids dissolved in chloroform were dried under high vacuum overnight and hydrated by Tris HCl 10 mm, pH 7.4 at room temperature. The lipid suspension was submitted to 5 freeze/thaw cycle (−183 °C/40 °C) and extruded 30 times through a polycarbonate membrane with 0.8 μm pores (Avanti Polar Lipids, Merck, Saint‐Quentin Fallavier, France) thanks to a mini‐extruder from Avanti Polar Lipids.

Turbidity and dynamic light scattering (DLS)

The turbidity of lipid vesicles was assessed by measuring the optical density at 436 nm (OD₄₃₆) at 22 °C thanks to a microplate reader Synergy 2 Biotek (Agilent, Les Ulis, France). After the measure of lipid vesicles absorbance at 0.5 mm in 95 μL, 5 μL of peptides or isopropanol were added at the right concentration. Measurements were performed every 3 min with 5 s of shaking before measurement. The size distribution of lipid vesicles was determined using a Malvern Zetasizer Nano series instrument (Malvern Instruments, Palaiseau, France). LUVs suspensions of 0.5 mm were incubated with the adequate concentration of peptides for 5 min at 22 °C. Quadruplicate measurements of 30 s were performed for each sample.

¹H NMR measurements

Lipid vesicles were prepared as described above and dissolved at 1 mm in 10% D₂O and Tris–HCl 10 mm pH 7.4. Peptides were resuspended in H₂O and added to lipid vesicles at the right concentration just before the spectra acquisition. All spectra were recorded using the Bruker Topspin software on a Bruker Avance 500 spectrometer equipped with a 5 mm TCl CryoProbe (¹H, ¹³C, ¹⁵N) (Bio‐RMN PRISM core facility, UAR 3480 CNRS – US18 Inserm, Univ Rennes, France). 1D spectra were recorded at 298 K with 128 scans. Water suppression was performed using pulse field gradient and excitation sculpting. Additional Tris buffer resonance signal suppression was made by presaturation during the relaxation delay.

Calcein leakage

Lipid vesicles containing calcein were prepared by solvent evaporation followed by extrusion. Briefly, lipids dissolved in chloroform were dried under high vacuum overnight and hydrated in Tris–HCl 10 mm containing 70 mm of calcein and 1 mm of EDTA at 40 °C for 1 h with regular vortex. Multilamellar vesicles were submitted to five freeze/thaw cycle (−180 °C/40 °C) and extruded 30 times through a polycarbonate membrane with 0.8 μm pores thanks to a mini‐extruder from Avanti Polar Lipids (Avanti Polar Lipids, Merck, Saint‐Quentin Fallavier, France). Lipid vesicles suspension was then separated from free dye by size exclusion chromatography using a Sephadex G‐50 column. The lipid content of the calcein‐loaded lipid vesicles fraction was estimated by ¹H NMR, by comparing the integrals of the fatty acid methyl signal (3H) and the tris methylene peak (6H). For permeabilization measurements, 30 μm of lipid vesicles were incubated in a microplate reader in a Synergy 2 Microplate reader (Agilent, Les Ulis, France) for 10 min with regular fluorescence measurements with excitation and emission wavelength at respectively 485 nm and 538 nm. Then, peptides or the equivalent quantity of ethanol were added at the right concentration and the fluorescence was followed for 15 min before the addition of 1% Triton X‐100. Measurements were performed at 22 °C. The percentage leakage for each sample was calculated using:

$$\%\text{leakage}=\frac{\left({\mathit{F}}_{15\min }-{\mathit{F}}_0\right)}{\left({\mathit{F}}_{\text{total}}-{\mathit{F}}_0\right)}$$

where F ₀ is the initial intensity of fluorescence before peptides addition, F _15min is the intensity of fluorescence after 15 min of peptides incubation and F _total is the intensity of fluorescence after Triton X‐100 addition.

Protoplasts permeabilization or lysis measurements

After 3 h of growth in TSB medium, 2 mL of cultures was centrifuged at 5000  g for 3 min, washed in PBS three times, and resuspended in digestion buffer (Tris–HCl 50 mm, NaCl 150 mm, Sucrose 750 mm pH 7.5) at OD₆₀₀ = 1. The culture was then incubated with 34 μg·mL⁻¹ lysostaphin to digest cell wall and 16 μg·mL⁻¹ DNase I for 1 h at 37 °C. Then, protoplasts were centrifuged at 9300 g for 15 min and resuspend in 500 μL of fresh digestion buffer. The efficiency of protoplast preparation was confirmed by plating 100 μL of the preparation in BHI plates; no colony was growing in the following 20 h at 37 °C. To assess the lysis and permeabilization of protoplasts after peptides addition, protoplasts were diluted in digestion buffer to obtain an OD₆₀₀ = 0.13 and incubated with 5 μm SYTOX Green for 5 min in the dark at 37 °C. 100 μL of solution was loaded in a microplate, and measurements of OD₆₀₀ and fluorescence (λ _exc = 488 nm, λ _em = 522 nm) were performed for 5 min with a Synergy 2 Biotek microplate reader (Agilent, Les Ulis, France). SprA1 or SprA2 peptides, at the appropriate concentrations, were then added and measurements of OD₆₀₀ and fluorescence (λ _exc = 488 nm, λ _em = 522 nm) were performed for 30 min before adding Triton‐X100 at a final concentration of 0.5%. As a control, the same experiment was done with whole cells. After 3 h of growth in TSB medium, 450 μL of culture was washed three times and resuspended in digestion buffer at OD₆₀₀ = 1. The cultures were then incubated with 5 μm SYTOX Green for 5 min in the dark at 37 °C. 100 μL of solution was loaded in a microplate and measurements of OD₆₀₀ and fluorescence (λ _exc = 488 nm, λ _em = 522 nm) were performed for 5 min with a Synergy 2 Biotek microplate reader (Agilent, Les Ulis, France).

Conflict of interest

The authors declare no conflict of interest.

Author contributions

LF contributed to the design of the experiments, conducted the experiments, performed their analysis, and wrote the manuscript. NA and CO performed the western blot experiments and northern blot experiment on HG003 pCN35 strains. YA, AR and AC reviewed the manuscript. IN performed peptides synthesis. SD carried out DNA sequencing. M‐LP‐M, SC, and AB, supervised the project, contributed to the design of the experiments, analyzed the data, and wrote the manuscript.

Peer review

The peer review history for this article is available at https://www.webofscience.com/api/gateway/wos/peer‐review/10.1111/febs.70001.

Supporting information

Table S1. Strains used in this study.

Table S2. Plasmids used in this study.

Table S3. DNA primers used in this study.

Table S4. Raw data used for this study.

Data availability statement

The data that supports the findings of this study are available in the Table S4.