The following S. aureus strains were used in this study: USA 300 strain LAC (Agr-QS subgroup I) and P3-luc (agr:P3-luc) (Nakamura et al., 2013); M1K051 (Agr-QS subgroup I, MSSA) and M1K155 (Agr-QS subgroup II, MSSA) (Nakamura et al., 2020); CN02 (Agr-QS subgroup III, MRSA) was isolated at Chiba University Hospital, Japan. S. aureus was cultured in Tryptic soy broth (TSB) (Millipore, USA) overnight at 37°C with shaking before use in the experiments. Frozen S. aureus stocks were maintained at −80°C in TSB supplemented with 40% glycerol. When counting colony-forming unit (CFU), samples were serially diluted in phosphate-buffered saline and plated on TSB agar plates. CFU counting was performed 18–24 h later. Escherichia coli DH5αZ1 was grown in Luria-Bertani (LB) medium. When introducing a plasmid, the LB medium was supplemented with 100 μg/mL ampicillin (Wako, Japan).

A luciferin-based, high-throughput assay was developed to screen 577 compounds extracted from natural sources in our laboratory (Arai et al., 2021). Among the 577 compounds screened, 499 were isolated from herbal medicines and 79 were originally from microorganisms. Overnight cultures of the agr:P3-luc strain were diluted 1:100 into TSB and incubated at 37°C for 4 h with shaking. Cultures were diluted with phosphate-buffered saline and transferred to a 96-well plate (Corning, USA) to obtain a final bacterial concentration of 2 × 10⁷ CFU/well. In experiments investigating the competitive effect between AIP-1 and compounds, 500 μg/L AIP-1 was added. Dimethyl sulfoxide (DMSO, 1%) was added as a control. The culture was incubated at 37°C for 24 h without shaking after which luminescence and an optical density at 600 nm (OD₆₀₀) were measured using Spark (TECAN, Switzerland).

Cultures were grown overnight growth and then diluted 1:1000 in TSB with 100 μM compound or 1% DMSO and incubated at 37°C for 8 h in a static state. The culture was then added to an equal volume of buffer (62.5 μM ethylenediaminetetraacetic acid, 100 μM NaCl, 1% sodium dodecyl sulfate) on a heat block at 95°C and heated for 2 min. Then, the mixture was reacted with an equal volume of phenol: chloroform: isoamyl alcohol (125:24:1) (Sigma, USA) and heated at 95°C for 5 min. The mixture was centrifuged, and the supernatant was used as the lysis sample. RNA was extracted using the Direct-zol RNA Miniprep kit (Zymo, USA), and cDNA was prepared using a PrimeScript™ RT Master Mix (Takara Bio, Japan). Quantitative PCR was performed using Brilliant III Ultra-Fast SYBR^® Green QPCR Master Mixes (Agilent, USA) and AriaMx Real-Time PCR System (Agilent, USA). The sequence of primers used for qPCR is listed in Supplementary Table S1; gyrB-F and gyrB-R for gyrB (Seidl et al., 2011), RNAIII-F and RNAIII-R for RNAIII (Seidl et al., 2011), agrA-F and agrA-R for agrA (Wang et al., 2021), hla-F and hla-R for hla (Jiang et al., 2023), and psma-F and psma-R for psm α (Jiang et al., 2018). Gene expression was normalized against that of gyrB, and relative expression was calculated using the 2^−ΔΔCT method (Livak and Schmittgen, 2001).

The hemolytic toxin assay was performed as previously described, with modifications (Bernheimer, 1988). Briefly, overnight cultures were diluted 1:1000 and incubated in TSB at 37°C with 100 μM compound or 1% DMSO for 4, 8, and 24 h. The cultures were centrifuged, and two-fold serial dilutions of the supernatant were prepared. Rabbit red blood cell solution (Japan BioSerum, Japan) was adjusted to provide an absorbance of 0.80 at 600 nm (A₆₀₀) when completely lysed with the addition of 2.5 mg/mL saponin. Two-fold serial dilutions of the supernatant and rabbit red cell solutions were incubated in a 96-well plate at 37°C for 30 min and centrifuged at 2000 g for 3 min. The degree of cell lysis was determined by measuring the A₆₀₀ of the supernatant. Data were fitted to a four-parameter logistic curve and expressed as HA₅₀ through nonlinear regression analysis, indicating the inverse of the dilution required for 50% complete lysis.

Cultures grown overnight were diluted 1:100 in Mueller Hinton Broth (MHB; BD Biosciences, USA) and incubated at 37°C with shaking until an OD₆₀₀ of approximately 0.5 was obtained. Following dilution with MHB, cultures were placed in a 96-well plate with compounds diluted in a two-fold series. This resulted in a final bacterial density of 2.5 × 10⁴ CFU/well. The culture was incubated at 37°C for 24 h without shaking, and bacterial growth was assessed by measuring the OD₆₂₀ using Infinite^® 200 PRO (TECAN, Switzerland).

Plasmid pTKY1297, encoding N-terminally histidine-tagged agrAc, was constructed by PCR amplification of a 313-bp BamHI–HindIII fragment carrying agrAc from the LAC strain using the primers BamHI-AgrA-F and HindIII-AgrA-R (Supplementary Table S1). The fragment was then cloned into the pUHE212-1 vector (Gamer et al., 1992). To purify N-terminally His-tagged AgrAc, 0.5 L of an E. coli DH5αZ1 derivative culture harboring pTKY1297 culture was incubated at 37°C for 2 h, and then isopropyl β-D-1-thiogalactopyranoside was added to obtain a final concentration of 1 mM for 3 h. The cells were collected by centrifugation and frozen at −80°C. Then, cells were suspended in buffer A (50 mM Tris-HCl pH 7.5, 300 mM NaCl, 10% glycerol, 0.1% Triton X-100) containing 10 mg/mL lysozyme (Sigma, USA) and sonicated on ice. The disrupted cells were centrifuged and the supernatant was incubated with Ni²⁺-NTA agarose (Qiagen, Germany) for 30 min. The AgrAc fusion protein was eluted using buffer A containing 500 mM imidazole. The protein solution was then dialyzed in buffer B (50 mM Tris-HCl pH 7.0, 50 mM NaCl, 1 mM dithiothreitol, 10% glycerol, 0.1% Triton X-100) using a Slide-A-Lyzer Dialysis Cassette (Thermo Scientific, USA) for 2 h. After replacing buffer B, the solution was dialyzed overnight, and the remaining solution was used as the purified AgrAc fusion protein.

To investigate the in vitro binding of AgrAc protein to DNA, an electrophoretic mobility shift assay (EMSA) was performed as previously described, with modifications (Sun et al., 2012). Briefly, 100- and 19-bp DNA fragments containing the P3 promoter and LytTR domain were synthesized and the 5’ end was labeled with Cy5 (Eurofins, Japan). DNA fragments were obtained by incubating with oligomer sets, Cy5-agrP3-F and Cy5-agrP3-R, and Cy5-LytTR-F and Cy5-LytTR-R, at 95°C for 5 min. The sequence of oligomers is given in Supplementary Table S1. Physalin H was added at concentrations of 31.25, 62.5, 125, 250, and 500 μM. Other physalins were added at a concentration of 500 μM. Then, 140 nM purified AgrAc fusion protein was added to 10 μL buffer (10 mM Tris-HCl pH 7.4, 50 mM KCl, 5 mM MgCl₂, 20% glycerol) and incubated at 25°C for 20 min. The mixture was incubated for a further 20 min after adding 100 nM of the DNA fragment. The reaction mixture was subjected to electrophoresis on a native 8% polyacrylamide gel in 0.5× TBE buffer. Labeled DNAs were detected using a Typhoon FLA9000 Photoimager (GE Healthcare, USA).

In silico docking simulations were performed using AutoDock Vina (Trott and Olson, 2010). Three-dimensional (3D) structures of compounds were prepared using Avogadro (Hanwell et al., 2012) and docked to the B chain of the crystal structure of AgrAc, which is deposited in the Protein Data Bank (PDB) under the accession number 4G4K (Leonard et al., 2012). The grid box was restricted to the DNA-binding region of AgrAc. Since initial simulations suggested that the compound binds to a pocket consisting of H169, N185, and N201, additional calculations were performed with a modified grid box and more flexible side chains near the ligand.

Molecular dynamics (MD) simulations of AgrA and compounds utilized the Desmond simulation package within the Schrödinger suite 2020-4 (Bowers et al., 2006). Before simulations, AgrA was prepared using the Protein Preparation Wizard in Desmond. An orthorhombic water box was established using the single-point charge (SPC) model (Zielkiewicz, 2006). Counterions were added to neutralize the solvent system, displacing water molecules and equalizing the net charge of the system to minimize its energy. Upon relaxation of the model system, simulations were initiated and performed at 300 K in an NPT (constant particle number, pressure, and temperature) ensemble. The simulations were conducted over 100 ns, with a recording interval of 20 ps for total energy. Interactions between compounds and AgrA were analyzed utilizing Desmond’s Simulation Interaction Diagram. The root mean square derivation (RMSD) for AgrA in the presence of compounds was calculated to examine the structural stability and conformational changes of the AgrA–physalin complexes. Root means square fluctuation (RMSF) analysis was performed to evaluate the conformational and local modifications resulting in physalin binding produced in the amino acid residues of AgrA.

Data were statistically analyzed using python 3.7. Data were analyzed using the two-tailed Student’s t-test. ns is used to denote values that are not significantly different.

S. aureus strain LAC, the epidemic methicillin-resistant USA300 clone, Agr-QS subgroup I, the predominant subgroup among all subgroups of Agr-QS, has been widely studied (Wang et al., 2007). First, we screened 577 compounds that inhibit agr:P3 activation using a reporter strain from S. aureus LAC, in which P3 drives the production of luminescence (Nakamura et al., 2013). To identify compounds that could inhibit Agr-QS expression without repressing bacterial growth, 100 μM of each compound was added to 2 × 10⁷ CFU/mL of bacteria and cultured for 24 h (Supplementary Figure S1A). Although 85 compounds generated luminescence levels less than 1% compared to those obtained with 1% DMSO, 81 compounds were considered to induce growth inhibition because they resulted in an OD₆₀₀ of 90% or lower than that obtained with DMSO (Supplementary Figure S1B). Overall, four compounds generated OD₆₀₀ of >90%. The four compounds were physalin H isolated from Solanum nigrum (Arai et al., 2014) (Figure 1A), lasidiol p-methoxybenzoate isolated from Xanthium strumarium (Karmakar et al., 2015), ovatodiolide isolated from Hyptis suaveolens (Arai et al., 2015), and ceanothic acid isolated from Zizyphus cambodiana (Arai et al., 2008), suggesting that these compounds can suppress Agr-QS in MRSA without repressing bacterial growth.

To assess the specificity of Agr-QS suppression by physalin H, agr:P3 activity was examined in the presence of serial concentrations of physalin H (Figure 1B). Regardless of the concentration of physalin H, the OD₆₀₀ and CFU values of culture after 24 h were comparable to those with 1% DMSO (Supplementary Figure S2A). A gradual concentration-dependent decrease in agr:P3 activity was observed with the addition of physalin H at concentrations of 5 μM and above, and no activity was observed when physalin H was added at 100 μM and above. Based on this, the IC₅₀ was calculated as 20.4 μM. Agr-QS activity increases during the exponential growth phase, inducing the expression of RNAIII and agrA (Novick et al., 1993). There was a marked decrease in the expression of RNAIII and agrA in LAC cells following 8 h culture with 100 μM physalin H (Figure 1C). RNAIII activates the expression of hla, encoding the hemolytic toxin, and phosphorylated AgrA activates the expression of psm α and psm β, encoding other hemolytic toxins (Queck et al., 2008). Additionally, the expression of hla and RNAIII decreased following 8 h culture of LAC cells with 100 μM physalin H (Figure 1C). Hemolytic toxins, such as Hla and phenol-soluble modulins (PSMs), are secreted by bacterial cells (Patel1 and Rawat2 (2023)). To examine the effect of physalin H on toxin production, hemolytic activity in the supernatant from LAC cells was determined (Figure 1D). Hemolytic activity in supernatants was repressed after 4, 8, and 24 h culture with physalin H compared with that in supernatants from culture with DMSO. Physalin H has also been reported to possess leishmanicidal activity (Choudhary et al., 2006). When physalin H was added to cultures with a bacterial count of ∼ 2.5 × 10⁵ CFU/mL and incubated for 24 h, high concentrations (200 μM) slightly inhibited bacterial growth (Figure 1E; Supplementary Figure S2B). These results indicated that physalin H inhibits Agr-QS expression in MRSA, suppressing the production of virulence factors.

To elucidate the target of physalin H in Agr-QS suppression, AIP-1 was added and agr:P3 activity was determined (Figure 1B). The addition of AIP-1 had no effect on the physalin H–dependent inhibition of P3 activity (IC₅₀ = 17.1 μM). Furthermore, reduced RNAIII expression level was observed in strains other than LAC, including M1K051 (Agr-QS subgroup I) and M1K155 (Agr-QS subgroup II) isolated from human skin, and CN02 (Agr-QS subgroup III) isolated as a nosocomial infection when cultured in the presence of 100 μM physalin H (Figure 1F). These results suggested that physalin H suppresses the function of AgrA in S. aureus rather than inhibiting the interaction between AIP and AgrC.

In addition to physalin H, eight physalins, PSs2–9, with different structures were isolated from P. minima; there were no significant differences in the physicochemical properties of these physalins (Supplementary Figure S3 ; Supplementary Table S2). These physalins are composed of eight rings, A–H, and only a small part of the difference in the structures between these physalins was in their A and B rings (Hosoya et al., 2008; Manome et al., 2023). The effect of PSs2–9 on Agr-QS suppression was investigated to clarify which substitution of the physalin structure is required for Agr-QS inhibition. Each physalin (100 μM) was added and the growth rate and agr:P3 activity were examined after 24 h (Figure 2A). Consequently, agr:P3 activity was almost completely absent in cells treated with PSs2 (physalin F), 3 (physalin B), and four (isophysalin B); however, agr:P3 activity remained at around 49%–61% with PSs5–9. When 100 μM PS2 was added, proliferation was about 80% compared with when 1% DMSO was added. However, like physalin H, treatment with PS2 also decreased agr:P3 activity in a concentration-dependent manner, even at 50 μM, which did not repress growth (IC₅₀ = 7.22 μM). A concentration-dependent decrease in agr:P3 activity was also observed following treatment with PSs3 and 4, with IC₅₀ values of 9.22 and 11.6 μM, respectively (Figure 2B). To confirm the effect of PSs2–9 on RNAIII expression in LAC cells, we examined RNAIII expression in LAC cells cultured for 8 h with 50 μM PS2 or 100 μM PSs3–9. The degree of RNAIII expression with PSs2, 3, and four was significantly decreased, whereas the degree of RNAIII expression with PSs5–9 was the same as that with DMSO (Figure 2C). The hemolytic activity of the 24-h culture supernatant following treatment with PSs3 and four was significantly reduced; however, that of culture supernatants following treatment with other physalins, including PS2, was similar to that of the culture with DMSO (Figure 2D). These results suggested that PSs3 and 4, in addition to physalin H, could repress virulence by inhibiting Agr-QS expression. Physalin D possesses antimicrobial activity against S. aureus with an MIC of 32 μg/mL (62.7 μM) (Helvaci et al., 2010). However, the MIC of PS5, corresponding to physalin D, against LAC exceeded 200 μM (Figure 2E). With PS2, weak growth inhibition was observed at 100 μM, and strong growth inhibition was observed at 200 μM. With PS3, weak growth inhibition was observed at 200 μM. No growth inhibition was observed with other PSs, even when 200 μM was added (Figure 2E).

AgrA binds to the P2 and P3 regions in the agr locus via its C-terminal DNA-binding domain (residues 138–238), termed AgrAc (Sidote et al., 2008). There is a high-affinity LytTR domain–binding site and a low-affinity binding site in the P3 region (Koenig et al., 2004). Therefore, to clarify whether physalin H inhibits the binding of AgrA to the P3 region, the effect of physalin H on the in vitro interaction between AgrAc and P3 DNA was investigated via EMSA (Figure 3A). The P3 region contains two binding sites for AgrAc; therefore, incubating purified N-terminally histidine-tagged AgrAc (His⁶-AgrAc) with a 100-bp oligonucleotide of the P3 region shifted the electrophoretic migration of Cy5-labeled nucleotides. DNA migration gradually decreased as the concentration of physalin H increased, and no DNA migration was observed following the addition of 500 μM physalin H. A 19-bp oligonucleotide containing the 9-bp consensus LytTR-binding sequence flanked by five bp on each side is sufficient for a stable interaction between AgrAc and DNA oligonucleotide (Sidote et al., 2008). DNA migration also shifted when 19-bp Cy5-labelled oligonucleotide and His⁶-AgrAc were mixed; however, this was lost following the addition of 500 μM physalin H (Figure 3B). Thus, physalin H blocks the stable interaction between AgrA and agr promoter regions.

To investigate whether differences in the Agr-QS inhibitory activity of physalins were due to differences in the inhibition of the AgrA–DNA interaction, we examined the binding between His⁶-AgrAc and 100-bp oligonucleotide with the addition of 500 μM of each physalin. All physalins inhibited the binding of AgrAc to DNA in vitro (Figure 3C).

The 9-bp consensus sequence interacts with AgrAc, and several bases make specific contacts with residues H169, N201, and R233 in AgrAc. The region flanking the 9-bp consensus sequence is involved in nonspecific DNA backbone interactions with AgrAc and appears to contribute to stability rather than specificity (Sidote et al., 2008). An in silico docking analysis between AgrAc and physalins (Figure 4) revealed three states where physalins interact, although the docking scores were similar, ranging between −10.3 and −8.9 kcal/mol (Table 1). Physalin H, with a docking score of −10.3 kcal/mol, formed a hydrogen bond with N201 of AgrAc through the carbonyl oxygen at C-18 in the E ring, and with R198 through the carbonyl oxygen at C-1 in the A ring (Figures 4A–C). Moreover, the hydroxyl group at C-13 may also interact with R198 through a hydrogen bond. Although the atoms used for hydrogen bonding between physalin H and AgrAc are common to all physalins, the entire ring structure of PSs2–4 interacted with AgrAc at the same location and in the same direction as physalin H (Figure 4B). However, the sites in PSs2–4 and the residues in AgrAc that form hydrogen bonds differed from those for physalin H (Table 1). The interaction of physalin H and PSs2–4 with AgrAc was designated as docking state A.

To enable PSs5 and six to interact with AgrAc, termed docking state B, PSs5 and six were placed in a similar position to that of physalins in docking state A by forming a hydrogen bond with R198 and N201 of AgrAc (the same as that formed by physalin H). However, the direction of the ring was opposite to that with physalins in docking state A because PSs5 and six form a hydrogen bond with N201 of AgrAc through the carbonyl oxygen at C-1, and with R198 through the carbonyl oxygen at C-18 (Figures 4D,E; Table 1). The docking simulation between PS7 and AgrAc suggested that PS7 located the docking state B position, although it did not form hydrogen bonds with any residues of AgrAc. PSs8 and nine interacted with AgrAc, termed docking state C, at a different position to that of physalins in docking states A and B. Furthermore, hydrogen bonds formed with the amino acid residues of AgrAc also differed (Figures 4F,G; Table 1). The carbonyl oxygen at C-1 in PSs8 and nine formed a hydrogen bond with H169 of AgrAc (Figure 4G). This may be due to differences in the binding state with AgrA; although Agr-QS inhibition was observed in physalins in docking state A, there was minimal Agr-QS inhibition for physalins in docking states B and C.

Although the positions of hydrogen bonds formed between the amino acid residues of AgrA and physalins in docked states A, B, and C differed, in each state, physalin interacted with amino acid residues important for AgrA binding with DNA (Figure 4; Table 1). To investigate the stability of the interaction between AgrA and physalins in docking states A, B, and C, 100-ns MD simulations were performed with AgrA and physalin H, PS5, and PS9 (Figure 5). First, we calculated the RMSD in the presence of physalins (Figures 5A–C). The protein RMSD of the AgrA–physalin H complex fluctuated up to around 2.0 Å for 10 ns and then to around 3.8 Å, with an average value of 2.96 Å. The ligand RMSD value also fluctuated up to 10 ns and then stabilized around 5.0 Å with an average value of 4.09 Å. This similar RMSD pattern suggested that physalin H stably binds to the binding pocket of AgrA. In contrast, the protein RMSD for AgrA–PS5 was similar to that obtained for physalin H; however, the ligand RMSD fluctuated up to 15 ns and then stabilized around 9.5 Å. The average values obtained for protein and ligand were 3.02 and 9.17 Å, respectively, indicating that the stability of PS5 in the docking pocket of AgrA was less than that of physalin H. For AgrA–PS9, the ligand RMSD was unstable and highly variable with an average value of 8.28 Å, and the protein RMSD stabilized at approximately 7 Å after 50 ns with an average value of 4.66 Å.

The RMSF plots for physalin H and PS5 were highly similar, with average RMSF values of 1.39 and 1.58 Å, respectively (Figures 5D,E). In the binding between AgrA and both physalins, although the residue index around 90–100 fluctuated, this position coincided with the boundary between the N- and C-terminal domains of AgrA. The AgrA region bound by both physalins was concentrated at a residue index around 130–150. In the binding between AgrA and PS9, although the binding sites of physalin were gathered in the same residue index, the PMSF plots were not stable compared with those for the other two physalins with an average RMSF value of 3.17 Å (Figure 5F).

AgrA–physalin interactions were monitored throughout the 100-ns simulation. As shown in Figure 5G, physalin H interacted with residues N185, L186, and N201 for >90% of the simulation period. Among interactions with N185 and L186, hydrogen bonds were maintained for more than 95% of the simulation period, and water bridges and hydrophobic contacts appeared partially. The interaction with N201 was present for about 40% of the simulation period through a hydrogen bond and for about 60% of the simulation period through a water bridge. The S202 residue interacted for about 40% of the simulation period, and of the 40%, 80% was through a hydrogen bond and the rest was through a water bridge. The L171 residue interacted for about 40% of the simulation period, and this interaction was a hydrophobic contact. Other than L237 and I238, the interacting residues were located in the DNA-binding region. In contrast, the number and position of amino acid residues in the DNA-binding region on AgrA with which the three physalins interacted were highly similar; however, unlike with physalin H, no residues continued to interact with PSs5 and nine in the 100-ns simulation, except for the interactions of PS5 with N185 and L186 (Figures 5H, I). The interaction of PS5 with L186 maintained a hydrogen bond for about 80% of the simulation period. The interaction between PS5 and N185 was present for about 90% of the simulation period, of which 30% was through hydrogen bonds and the rest through water bridge. These results suggest that the interaction between physalin H and AgrA is more stable than that between PSs5 and 9.

Snapshots of AgrA–physalin H every 20 ns in the 100-ns simulation revealed how the interaction between physalin H and the amino acid residues of AgrA changes ( Figure 6A; Table 2). At 0 ns, hydrogen bonds were formed between R198 of AgrA and the carbonyl oxygen at C-1 of physalin H, and between N201 of AgrA and the hydroxyl group at C-13 of physalin H, similar to the results obtained in the in silico docking simulation (Figure 4A). In addition, the hydroxyl group at C-6 of the B ring in physalin H interacted with N185 of AgrA through a hydrogen bond. At 20 ns, the two former hydrogen bonds were lost; however, the hydrogen bond between N185 and the hydroxyl group at C-6 remained, and an additional hydrogen bond was formed between L186 of AgrA and the carbonyl oxygen at C-15 of physalin H. Hydrogen bonds between N185 and the hydroxyl group at C-6, and between L186 and the carbonyl oxygen at C-15 were observed up to 100 ns. Hydrogen bonds between N201 and the carbonyl oxygen at C-1 were observed at 40, 60, and 100 ns. Furthermore, MD analysis of AgrA with PSs2–4, which have reduced Agr-QS activity in addition to physalin H, suggested that they continued to interact with L186 of AgrA and the carboxyl oxygen at C-15 of physalins through hydrogen bonds in the 100-ns simulation (Table 2; Supplementary Figure S4). In addition, N201 of AgrA interacted with the carboxyl oxygen at C-1 several times during the simulation period. These results suggest that the hydrogen bond formed between L186 and the carbonyl oxygen at C-15 acts as an anchor, sustaining the interaction between N201 and the carbonyl oxygen at C-1 of physalin H, PSs2, 3, and 4. It is possible that these interactions allow physalin to fit stably into the DNA-binding site of AgrA, thereby preventing AgrA from binding to the agr promoter. In contrast, PS5 formed a hydrogen bond with L186 at the carbonyl oxygen at C-26 in the G ring for 20–80 ns; however, this interaction did not allow PS5 to remain in the DNA-binding pocket of AgrA. Rather, it caused the entire structure of physalin to be pushed outside (Figure 6B).