Clinical S. aureus strains were characterized by Gram staining and the ability to produce coagulase and the clumping factor using Slidex Staph Plus (bioMèrieux, France). All isolates and reference strains used in the study are listed and described in Table 1. Bacterial cultures were grown aerobically in a nutrient trypticase soy broth TSB (bioMèrieux, France) at 37°C with shaking (150 rpm). SA147 cultures were supplemented with 100 μg/ml erythromycin.

The photosensitizers (PSs) used in the study were protoporphyrin IX diarginate (PPArg₂) (delivered by the Institute of Optoelectronics, Military University of Technology, Warsaw, Poland), toluidine blue O (TBO), 5.10.15.20-tetrakis(1-methyl-4-pyridinio)porphyrin tetra(p-toluenesulfonate) (TMPyP), zinc phthalocyanine (ZnPC) (Sigma-Aldrich, Munich, Germany) and N-methylpyrrolidinium fullerene iodide (fulleropyrrolidine, Full) (delivered by ProChimia, Sopot, Poland), and rose bengal (RB) Stock solutions were prepared as follows: 1 mM PPArg₂, TBO and TMPyP diluted in a sterile double distilled water and kept at −20°C; 10 μM RB diluted in a sterile double distilled water and kept at −20°C, 0.1 mM Full diluted in a mixture of DMSO and water (1:9 v/v) and kept at 4°C; and 1 μM ZnPC diluted in DMSO and kept at room temperature. 1.6-diphenyl-1.3.5-hexatriane (DPH) was purchased from Sigma-Aldrich (Germany). A 2 mM stock solution of this dye was prepared in tetrahydrofuran, and a 4 μM working solution was prepared by adding 100 μl to 50 ml of 0.05 M Tris-HCl (pH 7.6). Residual tetrahydrofuran was removed by gentle flushing with nitrogen (Bayer et al., 2000). A DPH working solution was kept in the dark at 4°C until use but no longer than 2 weeks.

Bacterial strains were cultured for 20–24 h and then diluted with a fresh broth to the density of 0.4 McFarland units (10⁷ CFU/ml). A total 100 μl of each culture was loaded into a 96-well plate and incubated in the dark at 37°C for 30 min, either with or without the addition of PS. The following concentrations of PSs were used: PPArg₂, TBO, TMPyP–20 μM; RB–0.1 μM; Full–1 μM; ZnPC–5 nM. Then, samples were irradiated with light doses ranging from 10 to 30 J/cm². For the irradiation procedure, LED lamps (SecureMedia, Poland), designed and produced for the Laboratory of Molecular Diagnostics, emitting incoherent red light (615–635 nm, maximal intensity at λ_max 627 nm) or green light (500–575 nm, λ_max 520 nm) were used. Technical parameters were set as follows: output power of 23.4 mW/cm² and energy dose of 10–30 J/cm². Aliquots incubated in the dark with and without PS served as dark controls. After irradiation with each 10 J/cm² light dose, 10 μl of the samples were taken to perform ten-fold serial dilutions in PBS, ranging from 10⁻¹ to 10⁻⁴. Ten microliter aliquots of each dilution were plated on TSA plates (bioMèrieux, France). After overnight incubation at 37°C, the colonies formed were counted, and the results were analyzed statistically. Each experiment was performed at least in triplicate.

Bacterial cultures (20–24 h) were suspended in PBS buffer to a density of 0.4 McFarland units. Bacterial suspensions containing 5 × 10⁷ CFU in 800 μl volume were centrifuged at 5000 × g for 5 min, washed and resuspended in 800 μl PBS. Suspensions were incubated with 20 μM PPArg₂ in the dark at 37°C for 30 min and then centrifuged at the described conditions. Absorbance of supernatants was measured at 408 nm in 96-well plates using an EnVision® Multilabel Plate Reader (Perkin-Elmer, USA), and the amount of residual PPArg₂ was calculated in reference to a calibration curve. The amount of photosensitizer accumulated by the cells was calculated by subtracting residual PPArg₂ concentrations from the initial concentration, that is 3.2 nmols per 10⁷ cells.

Bacterial DNA was isolated from 1 ml of the 24 h culture a using GeneJET™ Genomic DNA Purification Kit (Thermo Scientific, USA), according to the manufacturer's protocol. The DNA concentrations were measured using a NanoDrop™ ND-1000 (Thermo Scientific, USA). The primers used for gene amplification were as follows: for rsbU 5′-atctctgcagTTAATTTACTCTTTTTATAATC-3′ and 5′-atcgaattcaTGGAAGAATTTAAGCAAC-3′; for rsbV and rsbW in one product 5′-TTAGCTGATTTCGACTCTTTCG-3′ and 5′-atcgaattcATGAATCTTAATATAGAAACAACC-3′; for sigB 5′-atcgaattcaTGGCGAAAGAGTCGAAATCAGC-3′ and 5′-atctctgcagCTATTGATGTGCTGCTTCTTG-3′ (Senn et al., 2005). PCR was performed with 1 U of Taq polymerase (Sigma-Aldrich, Germany) and 100 ng of template DNA. PCR products were purified using a QIAquick™ PCR Purification Kit (Qiagen, The Netherlands), and DNA sequencing was accomplished from both ends by the Sanger dideoxy-chain termination method according to standard (500–800 nt reads) or extra-long run protocols (1000–1100 nt reads), depending on template length. Sequences were analyzed and aligned to the reference AP009351.1 sequence (S. aureus Newman) using BioEdit Sequence Alignment Editor 7.2.0 (Hall, 1999).

Overnight bacterial cultures were added to fresh broth in a 1:100 ratio and cultured to reach mid-logarithmic phase of growth (~2.5 h). For total bacterial RNA extraction, 500 μl of cell suspension of OD₆₀₀ 0.5 density (10⁸ CFU/ml) was mixed with 1 ml of RNAprotect® Bacteria Reagent (Qiagen, The Netherlands) according to the manufacturer's instructions. Bacterial cells were then lysed in 100 μl of buffer containing 20 mM Tris-HCl (pH 8.0), 2 mM EDTA, 1.2% Triton X-100 and 2 U lysostaphin (A&A Biotechnology, Poland) at 37°C for 30 min. Bacterial suspensions were further homogenized in MagNA Lyser instrument (Roche, France) with acid-washed glass beads (Sigma-Aldrich, Germany) for 3 min at 6000 rpm. RNA extraction was performed with an RNeasy® Mini Kit (Qiagen, The Netherlands) according to the manufacturer's protocol, with additional on-column DNase I treatment. The quality of total RNA was evaluated on a 1.5% agarose gel, and its concentration was measured using a NanoDrop™ ND-1000 (Thermo Scientific, USA). First strand cDNA synthesis was performed using a TranScriba Kit (A&A Biotechnology, Poland) with the addition of random hexamers and 100 ng total RNA. For qPCR, the asp23-specific primers 5′-AAAGCAAAACAAGCATACGACAATC-3′ and 5′-AGCGATACCAGCAATTTTTTCAAC-3′ and reference 16SrRNA specific primers 5′-TATGGAGGAACACCAGTGGCGAAG-3′ and 5′-TCATCGTTTACGGCGTGGACTACC-3′ were used (Ster et al., 2005). cDNA quantification was performed in 10 μl volume with SYBR Green I containing SG qPCR Master Mix (EURx, Poland), 0.25 μM of each primer and 1 μl of 1:10 diluted cDNA using a 480 Light Cycler® instrument (Roche, France). The following protocol was applied: 3 min of initial denaturation at 95°C and 45 cycles of denaturation at 95°C for 10 s, primer annealing at 62°C for 30 s and elongation at 72°C for 30 s, with melting curve analysis at a range of 60 to 90°C. Normalized values of asp23 expression in analyzed strains were calculated relative to SH1000 calibrator, based on threshold numbers. Each experiment was performed at least in triplicate.

We used RsbU amino acid S. aureus sequence deposited under the GenBank number: AP009351.1 to predict the homology model of this protein.

The main difficulty in predicting the structure is the lack of a single protein structure that has significant sequence identity with the target sequence in the PDB database (Protein Data Base). Therefore, multiple templates were used to improve the quality of the obtained target homology model. Up to now, no data describing the 3D structure of this protein was available. To uncover as much structural detail as possible, we conducted the following bioinformatic analyses: the search for structural homologs in the SWISS−PROT database [http://web.expasy.org/docs/swiss-prot_guideline.html (accessed 30.04.2015)] using BLAST [http://blast.ncbi.nlm.nih.gov/Blast.cgi (accessed 30.04.2015)]; an analysis of the secondary structure using PSIPRED [http://bioinf.cs.ucl.ac.uk/psipred/ (accessed 30.04.2015)], SSPRO (Cheng et al., 2005a) and DISEMBL (Linding et al., 2003); searching for unstructured protein fragments using DISOPRED (Jones and Cozzetto, 2015), DISEMBL (Linding et al., 2003) and DISPRO (Cheng et al., 2005b); the identification of transmembrane helices using TMHMM [http://www.cbs.dtu.dk/services/TMHMM-2.0/ (accessed 30.04.2015)], TOPRED (von Heijne, 1992), HMMTOP (Tusnády and Simon, 1998) and MEMSAT (Jones, 2007); prediction of regions with a coiled-coil structure using COILS (Lupas et al., 1991). Additionally, we searched for evolutionarily conserved domains using Reversed Position Specific Blast (RPS−BLAST) [http://www.ncbi.nlm.nih.gov/Structure/cdd/wrpsb.cgi (accessed 30.04.2015)]. The final homology model of the target sequence was assembled with MODELER (Eswar et al., 2006) software.

The bacterial cultures were grown for 24 h. Bacterial cells at 2 × 10⁹ (OD₆₀₀ 2.0 in 5 ml volume) were harvested at 4000 rpm at 4°C for 10 min and washed twice with double distilled water. After centrifugation, pellets were suspended in 1.5 ml of 99% methanol (POCH, Gliwice, Poland) and agitated for 2 h in the dark until bleached. Samples were centrifuged at 10,000 × g at 4°C for 15 min, and the absorbance of the supernatant was measured at 450 nm using a Novaspec II spectrophotometer (Pharmacia Biotech, USA). Each experiment was performed at least in triplicate.

Temperature-dependent membrane fluidity was quantified by measuring the fluorescence polarization anisotropy (r) of DPH, according to modified protocols described by Bayer and Voss (Bayer et al., 2000; Voss and Montville, 2014). DPH, a lipophilic fluorescent probe, preferentially localizes to hydrophobic (intrinsic) regions of cell membrane phospholipids and emits polarized light upon excitation. The value of fluorescence anisotropy (r) reflects unequal intensities of emitted light along different axes of polarization (vertical and horizontal) and therefore indicates the degree of a fluorophore's free movement that is dependent on cell membrane fluidity. Whole-cell suspensions of each bacterial strain were prepared with a density of 4.5 McFarland units (10⁸ CFU/ml) in TSB medium. Suspensions were pelleted by centrifugation (5000 × g, 15 min) and then resuspended in 500 μl of digestion buffer (20% [w/v] sucrose, 0.05 Tris-HCl [pH 7.6], 0.145 M NaCl). The bacterial cell wall was then digested with 0.8 U of lysostaphin (A&A Biotechnology, Poland) in the presence of 3 U of DNase I (EURx, Poland) for 1 h at 37°C (Bayer et al., 2000). Protoplasts were collected by centrifugation (10,000 rpm, 15 min) and resuspended in 200 μl of fresh digestion buffer. The adequacy of cell wall digestion was confirmed by Gram staining. For DPH labeling, protoplasts suspended in digestion buffer were mixed with DPH solution in a 1:1 ratio to obtain 2 μM final concentration and incubated in the dark at 30°C for 45 min. Spectrofluorimeter FP-8500 (JASCO, USA) coupled with Spectra Manager™ software was used for fluorescence intensity and anisotropy measurements. Analysis was carried out in a labeled cell suspension volume of 300 μl, agitated at 200 rpm, in a temperature gradient ranging from 20 to 40°C (ramping rate 1°C per 1 min). Above this temperature, a disruption of labeled protoplasts was observed using fluorescence microscopy. A blank measurement was recorded using the unlabeled cell suspension of each strain separately, at a single initial point, as no significant changes in background fluorescence intensity of unlabeled protoplasts were observed during the whole measurement. Measurement parameters were set up as follows: a vertically polarized excitation wavelength of 360 nm (bandwidth 5 nm) and an emission wavelength of 426 nm (bandwidth 10 nm) analyzed through a rotating polarizer. Signals were measured for 2 sec at each 2.5°C interval. A G factor of 1.2017 was used. Each experiment was performed at least in triplicate.

The results of photodynamic inactivation, carotenoid extraction, photosensitizer uptake and quantitative PCR are presented as the average of at least three independent experiments. Statistical significance was assessed using Student's t-distribution method. Grouping of fluorescence anisotropy data included analysis of correlation matrix within a whole temperature range followed by hierarchic clustering based on Ward agglomeration and Manhattan distance calculation methods using STATISTICA 10 software (StatSoft Inc. 2011, USA). Three biological replicates of each S. aureus strain were included separately into the analysis.

To assess the impact of sigB operon impairment on PACT outcome, we first tested the survival of an isogenic set of USA300 mutants treated with increasing light doses (10-30 J/cm²) and various photosensitizers. The results are presented in Table 2 as means of viability reduction for a discriminating light dose of 20 J/cm². Full spectrum of the applied light doses are presented in Supplementary Figure 1. A clear difference in vulnerability to PACT was observed between wild type USA300 and sigB operon mutants in the presence of PPArg₂, ZnPC and RB. Here, the applied treatment was inefficient toward the wild type USA300 but led to almost complete eradication of each σ^B-impaired mutant. By eradication, we considered a reduction of more than 5 log₁₀ units in the number of CFU; thus, the drop below a detection threshold of a method. In the presence of TBO and TMPyP, applied irradiation was ineffective toward both wild type and isogenic mutants. Fulleropyrrolidine, although more potent under the herein applied conditions, revealed no discriminative activity with regards to wild type and mutant strains. To verify this observation in a different genetic background, we used another set of strains: RN6390 and SH1000. In RN6390, the rsbU gene contains a natural 11-nt deletion, and this deletion is restored in strain SH1000 (Horsburgh et al., 2002). In this system, the lack of functional RsbU—a σ^B activator—led to a similar PACT outcome, as observed previously. We noticed effective killing of strain RN6390 upon PPArg₂, ZnPC and RB treatments and poor phototoxic activity toward strain SH1000 (Table 2).

Based on the presented results, we conclude that the sigB operon affects PACT efficacy with the use of PPArg₂, ZnPC and RB in S. aureus.

To analyze the correlation between PACT outcome and σ^B activity, clinical S. aureus isolates previously described as susceptible or insensitive to protoporphyrin IX-mediated PACT (Grinholc et al., 2008) were included in the analysis in a context of sigB operon integrity First, we characterized phototoxic activity of the three analyzed PSs (PPArg₂, ZnPC, and RB) toward these clinical isolates of S. aureus (Table 3). From the obtained data, it can clearly be seen that each of the tested clinical isolates is sensitive to at least two out of three analyzed PSs (Table 3).

We then sequenced the genes of the sigB operon (rsbU, rsbV, rsbW, sigB) and screened for mutations in these PACT-vulnerable isolates. Interestingly, all the identified mutations were localized to the rsbU gene (except for 146 strain, where another N61K substitution in RsbW was also identified). Each PACT-sensitive strain carried different types of variations, which were absent in the less sensitive to PACT strain 5491 (Table 4, Figure 1). The obtained results do not rule out other mutations outside sigB operon as associated with PACT-vulnerable phenotype. Interestingly, the applied irradiation conditions did not lead to PPArg₂-based eradication of strain 6452 in spite of an obvious aberration within the rsbU gene. This phenomenon can be connected to significantly lower PPArg₂ uptake compared to other analyzed strains (Supplementary Figure 2).

To date, neither the experimental nor predicted structure of RsbU from S. aureus has been published. The crystal structure of the RsbU ortholog in B. subtilis, which serves as a model of σ^B activation system, has been solved only for the N-terminal regulatory domain (Delumeau et al., 2004). To show the sites of detected mutations within the protein structure and assess their potential relevance, the theoretical structural model of the S. aureus RsbU monomer was created and presented with an indication of the identified changes (Figure 2). The results of the bioinformatic analyses allowed us to hypothesize that the 3D structure single-chain conformation of the target sequence consists of the two domains connected with one relatively long α-helix.

Furthermore, the N-terminal domain was found to be all-helical, while the C-terminal domain was predicted to accommodate an αβ-structure. For modeling of the 3D structure of RsbU, we employed two template structures. The first template was the protein structure with the PDB code 3T91 (chain A), in which the 242-residue fragment shared 17% amino acid sequence identity and 42% homologous residues with the target. This template served to predict the RsbU protein fragment from residue 82 to 330. The second template was PDB code 1W53 (chain A), in which the 84-residue fragment shared 29% amino acid residue sequence identity and 54% homologous residues with the target, and it predicted the RsbU protein fragment from residue 2 to 82. We did not predict the conformation of the first residue or the last three residues of the target sequence. The final 3D structure of the RsbU homology model assembled by MODELER (Eswar et al., 2006) consists of two domains: an N-terminal α-helical four helix bundle and C-terminal αββα arrangement, similar to PP2C-type phosphatases (Schweighofer et al., 2004).

The two flanking domains are connected with a long α-helix. Interestingly, both templates form dimers, which may be evidence of a tendency of the target protein structure to form dimers. Dimer formation has also been proposed for B. subtilis RsbU, suggesting the localization of dimerization-determining motifs to the N-terminal domain (Delumeau et al., 2004). From the present model, it can clearly be seen that four identified mutations localize in the C-terminal domain, two in the N-terminal domain and one in the linker helix in between the two domains, probably resulting in protein truncation and lack of a C-terminal regulatory domain.

To check if the identified mutations result in a σ^B-defective phenotype in clinical isolates, a functional test was employed. We analyzed the transcript levels of the membrane-anchored alkaline shock protein Asp23, which is exclusively controlled by σ^B factor (Gertz et al., 1999; Müller et al., 2014). The qPCR results were calculated relative to SH1000 transcript levels (Figure 3). Elevated mRNA levels were observed in the clinical isolate strain 5491 with respect to SH1000, in which no missense mutations were detected in the rsbU gene. Accordingly, a significant 82.5-fold decrease in asp23 transcript levels was observed in the RsbU-defective strain RN6390. Four out of six PACT-vulnerable isolates (473, 6987, 6452, 1764/p) showed decreased asp23 transcript levels, among which 6452 and 6987 revealed statistical significance (Figure 3). On the contrary, elevated transcript levels were observed in strains 469 and 146 compared to SH1000, indicating that the identified substitutions in RsbU (together with N61K substitution in RsbW in strain 146) had no diminishing effect on σ^B activity.

Methanol extracts from S. aureus contain a mixture of various pigments with absorbancemaximum around λ = 450 nm, ranging from a pale yellow 4.4′-diapo-ζ-carotene to a red 4.4′-diaponeurosporenal. These constitute the intermediate products of the staphyloxanthin (STX) synthesis pathway and are generally referred to as carotenoids (Marshall and Wilmoth, 1981). RsbU-activated σ^B drives the expression of membrane-localized STX in S. aureus (Giachino et al., 2001; Olivier et al., 2009). Therefore, we used STX levels as another functional test of σ^B activity. As a reference system, we used a subset of isogenic strains differing in pigmentation status (SA144, SA145, SA147; see Table 1). We measured carotenoid levels in all analyzed strains (Figure 4). Reference strains with an intact sigB operon, namely USA300 and SH1000, accumulated similar amounts of carotenoids as strain SA144. As expected, RN6390 with non-functional RsbU accumulated significantly less carotenoids than SA144. Similarly, each ΔrsbUVWsigB isogenic mutant accumulated significantly less carotenoids compared to reference SA144. The level of carotenoids in clinical isolates stayed in compliance with asp23 transcript-coupled σ^B activity. As expected, carotenoid synthesis was strongly inhibited in isolates 473, 6987, 6452, and 1764/p, where decreased activity of σ^B was observed (Figure 3). Extracts of strains 5491, 146, and 469 revealed similar absorbance values as the reference wild type SA144 strain, indicating relatively high carotenoid levels.

Carotenoid pigments are broad-range reactive oxygen species scavengers, showing the ability to neutralize free radicals as well as efficiently quench singlet oxygen (¹O₂) (Clauditz et al., 2006). To assess if the lack of pigmentation is crucial and sufficient for determining PACT outcome, we first assessed the phototoxic effect of PPArg₂, ZnPC and RB on reference strains SA144, SA145, and SA147, in which carotenoid production is regulated at the level of the crtOPQMN operon (Table 5). Notably, in all three cases of the studied PSs, the non-pigmented strain SA145 was the most vulnerable to PACT treatment compared to its pigmented counterparts (SA144, SA147). A significant bactericidal effect or total eradication was observed after RB and ZnPC treatments but not PPArg₂. Additionally, the ZnPC bactericidal effect on wild type strain SA144 was abolished in the carotenoid-overproducing strain SA147. A similar trend could be seen in the case of RB; however, the values of survival reduction for this PS were lower. Based on the obtained results, pigmentation levels correlated with photokilling efficacy with the three PSs used in our experimental conditions. In the case of clinical isolates, carotenoid levels correlated with RsbU-dependent σ^B activity and may have a vital contribution to PACT outcome. Nevertheless, pigment levels remained relatively high in the PACT-sensitive strains 146 and 469, with undisturbed σ^B activity, suggesting that other mechanisms must be involved.

Bacterial membrane fluidity, among other factors, such as protein content, also reflects changes in staphyloxanthin level (Mishra et al., 2011). Apart from providing antioxidant defense, the STX biosynthesis pathway shares similarities with that of human cholesterol, suggesting similar roles in maintaining cell membrane fluidity. We measured temperature-dependent fluorescence anisotropy (r) in DPH-labeled bacterial cell membranes. An inverse relationship occurs between measured DPH fluorescence anisotropy and membrane fluidity (Mishra et al., 2011). Statistical analysis of r-values revealed a strong dependence of all measurements within the studied temperature range, presenting the same trend in anisotropy drop upon heating (Supplementary Figure 3). Figure 5 presents the r-values for optimal growth temperature of S. aureus at 37.5°C. The statistical analysis resulted in distinguishing three groups of strains with distinct fluorescence anisotropy values (Supplementary Figure 3). Analysis of the strains SA144, SA145 and SA147 confirmed a pronounced difference between pigmented and non-pigmented cells. High r-values (low membrane fluidity) were observed for pigmented strains SA144 and SA147 (Figure 5, Group 1). The reference pair of SH1000 and RN6390 also revealed significantly different r-values; however, despite the similar carotenoid content in SH1000 and the strains ranked in the first group (SA144, SA147, 5491), the former was included in Group 2. This leads to the assumption that STX is not the only factor affecting overall membrane fluidity. The third group (Figure 5, Group 3) presenting the highest membrane fluidity included the strains SA145, RN6390 and 1764/p, which are characterized by poor pigmentation, and strain 146. In the case of strains 146 (Group 3) and 469 (Group 2), which presented neither the abolished σ^B activity nor significantly inhibited carotenoid accumulation, we suppose that a high membrane fluidity of unknown background is the factor sensitizing those isolates to photosensitizer action.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The reviewer EF and handling Editor declared their shared affiliation, and the handling Editor states that the process nevertheless met the standards of a fair and objective review.