Computational Exploration of the Self-Aggregation Mechanisms of Phenol-Soluble Modulins β1 and β2 in Staphylococcus aureus Biofilms

Abstract

The formation of functional bacterial amyloids by phenol-soluble modulins (PSMs) in Staphylococcus aureus is a critical component of biofilm-associated infections, providing robust protective barriers against antimicrobial agents and immune defenses. Clarifying the molecular mechanisms of PSM self-assembly within the biofilm matrix is essential for developing strategies to disrupt biofilm integrity and combat biofilm-related infections. In this study, we analyzed the self-assembly dynamics of PSM-β1 and PSM-β2 by examining their folding and dimerization through long-timescale atomistic discrete molecular dynamics simulations. Our findings revealed that both peptides primarily adopt helical structures as monomers but shift to β-sheets upon dimerization. Monomeric state, PSM-β1 exhibited frequent transitions between helical and β-sheet forms, while PSM-β2 largely retained a helical structure. Upon dimerization, both peptides showed pronounced β-sheet formation around conserved C-terminal residues 21–44. Residues 21–33, largely unstructured as monomers, demonstrated strong tendencies for β-sheet formation and intermolecular interactions, underscoring their central role in the self-assembly of both peptides. Additionally, the PSM-β1 N-terminus formed β-sheets only when interacting with the C-terminus, whereas the PSM-β2 N-terminus remained helical and uninvolved in β-sheet formation. These distinct aggregation behaviors likely contribute to biofilm dynamics, with C-terminal regions facilitating biofilm formation and N-terminal regions influencing stability. Targeting residues 21–33 in PSM-β1 and PSM-β2 offers a promising therapeutic approach for disrupting biofilm integrity. This study advances our understanding of PSM-β1 and PSM-β2 self-assembly and presents new targets for drug design against biofilm-associated diseases.

Graphical Abstract

1. Introduction.

Amyloids—self-assembled from a range of peptides and proteins into highly stable, insoluble β-sheet structures—play significant roles in both health[1, 2] and disease[3–6], and show promise in biomedical and nanotechnology applications[7–9]. Pathological amyloids, such as amyloid-β (Aβ) and tau in Alzheimer’s disease (AD)[5, 6, 10, 11], α-synuclein in Parkinson’s disease (PD)[4, 12, 13], and human islet amyloid polypeptide (hIAPP) in type 2 diabetes (T2D)[6, 14, 15], contribute to neurodegeneration through irreversible aggregation that drives disease progression[16]. In contrast, functional amyloids are beneficial, performing essential biological roles across all life forms, from bacteria to humans[1, 17, 18]. These amyloids participate in processes like melanin production (e.g., Pmel17[19]), hormone storage (e.g., glucagon[20] and β-endorphin[21, 22]), and biofilm stabilization (e.g., Curli[23], Fap[24], and phenol-soluble modulins (PSMs)[25, 26] fibers in bacteria). In microorganisms, functional bacterial amyloids (FuBAs) are key components of biofilms, reinforcing the extracellular polymeric substance (EPS) matrix—composed of polysaccharides, proteins, and extracellular DNA—that enables bacterial adhesion, nutrient retention, and a protective barrier against antimicrobial agents and immune defenses[17, 26, 27]. This robust structure makes biofilm-associated infections, responsible for about 80% of chronic infections, exceptionally difficult to treat[28, 29]. Among these, methicillin-resistant Staphylococcus aureus (S. aureus) is particularly concerning, classified by the World Health Organization as a high-priority pathogen[30]. Understanding the molecular mechanisms of FuBA formation is crucial for developing targeted strategies to disrupt biofilms, paving the way for more effective treatments for persistent, biofilm-related infections.

The biofilm matrix of pathogenic S. aureus, a gram-positive bacterium, is reinforced by functional amyloids derived from a group of seven small peptides known PSMs[30, 31]. These PSMs are categorized by size into PSM-α, PSM-β, and δ-toxin, with distinct roles depending on their structural state[26]. In their soluble monomeric form, PSMs can disrupt host immune responses by recruiting, activating, and lysing human neutrophils while promoting biofilm dispersal[32, 33]. Conversely, when PSMs self-assemble into amyloid fibrils, they strengthen the biofilm matrix, enhancing its resistance to mechanical disruption and enzymatic degradation[32, 33]. The genes encoding these PSMs are highly conserved: PSM-α (α1–α4) is encoded by the PSM-α operon, PSM-β (β1 and β2) by the PSM-β operon, and δ-toxin is located within the RNAIII coding sequence[25, 26]. High expression levels of PSM-α peptides, which are approximately 20 amino acids long, correlate with increased virulence in methicillin-resistant S. aureus, particularly due to the potent cytotoxic effects of PSM-α3, which exhibit enhanced toxicity upon amyloid fibrillation[25, 31]. The longer PSM-β peptides (~44 residues) are essential for maintaining biofilm structural integrity[33–35]. Antibodies targeting PSM-β peptides effectively inhibited bacterial spread from indwelling medical devices, highlighting the crucial role of PSM-β peptides in regulating S. aureus biofilm formation[33, 34]. At moderate concentrations, PSM-β peptides promote biofilm formation, with PSM-β1 being more potent; however, at higher concentrations, they inhibit biofilm formation and induce detachment, with PSM-β2 demonstrating greater effectiveness[33]. Despite progress in understanding PSM amyloids in S. aureus biofilms, the molecular mechanisms driving their assembly from soluble monomers to fibrillar structures remain elusive. Additionally, the conformational shifts that enable monomeric PSMs to promote biofilm dispersal, while fibrillar forms reinforce the biofilm matrix[32, 33], are still unclear. Clarifying these processes is essential for developing strategies to combat biofilm-related infections.

The monomeric forms of PSM-β1 and PSM-β2 have been shown to adopt helical structures, as confirmed by synchrotron radiation circular dichroism (CD) and Fourier transform infrared (FTIR) spectroscopy[26, 36–38]. In CD spectral measurements, PSM-β2 exhibited a higher abundance of helical structures compared to PSM-β1, even though both peptides displayed pronounced unstructured characteristics in the absence of trifluoroethanol (TFE) [26, 38]. The atomic structure of PSM-β2 has been characterized using nuclear magnetic resonance (NMR) spectroscopy, revealing three helical regions spanning residues 5–15, 17–23, and 25–43[36]. In contrast, the atomistic structure of PSM-β1 remains elusive, likely due to the highly dynamic nature of its helical conformation. Upon aggregation, both PSM-β1 and PSM-β2 tend to transition to β-sheet-rich structures, influenced by environmental factors such as pH and temperature[25]. The spectral changes observed during the aggregation process of both peptides closely resemble those seen in disease-related amyloids, such as Aβ and hIAPP[10, 15]. Despite their sequence homology, PSM-β1 and PSM-β2 exhibit distinct properties[39, 40]. For instance, PSM-β2 is more sensitive to environmental factors than PSM-β1. Zahra et al. found that PSM-β1 aggregates in the presence of 1 mg/mL of human plasma protein fibrinogen, while PSM-β2 does not[40]. Additionally, acidic pH conditions have a more pronounced inhibitory effect on PSM-β2 fibrillation compared to PSM-β1[39]. While both PSM-β1 and PSM-β2 contribute to biofilm formation[29, 33], the conformational dynamics during their self-assembly, the mechanisms of aggregation, and the impact of sequence variations on their structural properties remain poorly understood.

In this study, we explored the conformational dynamics of PSM-β1 and PSM-β2 during self-aggregation, focusing on their folding and homodimerization through atomistic discrete molecular dynamics (DMD) simulations[41, 42]—a method renowned for its efficiency and accuracy in modeling amyloid formation[31, 43, 44]. Our simulations revealed significant structural distinctions between the two peptides. PSM-β1 displayed high structural flexibility, frequently transitioning between helical and β-sheet conformations, whereas PSM-β2 predominantly maintained a stable helical structure, especially within its N-terminal region. Upon dimerization, both peptides exhibited increased β-sheet content and a reduction in unstructured regions, with residues 21–33 showing the strongest β-sheet tendency and contributing the greatest number of interpeptide contacts, suggesting a pivotal role in driving dimerization and self-assembly. In PSM-β1 dimers, residues 21–33 supported additional β-sheet formation within residues 11–28 and 35–44, while in PSM-β2, these residues primarily facilitated β-sheet formation in the C-terminal residues 35–44, with the N-terminal residues 1–20 largely retaining its helical conformation. Overall, PSM-β1 exhibited a higher propensity for β-sheet formation than PSM-β2, highlighting distinct conformational dynamics in biofilm development. These findings underscore the critical role of residues 21–33 in stabilizing β-sheet structures through dimerization and suggest that targeting this conserved region could present a promising strategy to disrupt PSM-β1 and PSM-β2 aggregation, potentially inhibiting biofilm formation and offering new approaches for managing biofilm-associated infections.

2. Materials and methods

2.1. Molecular systems

The amino acid sequences of PSM-β1 (Uniport: A0A0H3KCA8) and PSM-β2 (Uniport: A0A0H3KSC6) used in the simulations are presented in Figure 1. Since the solution structure of the PSM-β1 monomer has not yet been experimentally determined, an AlphaFold2[45] prediction served as the initial model. For PSM-β2, the initial structure was obtained from the Protein Data Bank (PDB: 5KGZ), based on an NMR study conducted in an aqueous TFE medium[36]. To ensure comprehensive conformational sampling and explore the dynamics of PSM-β1 and PSM-β2 monomers, 50 independent discrete molecular dynamics (DMD) simulations were performed for each monomer, starting with randomized velocities. Each simulation ran for up to 600 ns to minimize biases arising from the initial configurations. Additionally, to investigate the conformational changes during the self-assembly of PSM-β1 and PSM-β2 dimers, two additional systems—each containing two PSM-β1 or two PSM-β2 peptides—were established. For each system, 50 independent simulations were conducted, with each simulation running up to 1000 ns. To avoid potential bias from initial configurations, each simulation started with distinct coordinates and velocities. In the dimerization simulations, peptides were randomly placed within a cubic simulation box with varying orientations, ensuring a minimum intermolecular distance of no less than 1.5 nm. Further details about each molecular system are consolidated in Table 1.

Figure 1. Amino acid sequences and initial structures of PSM-β1 and PSM-β2 monomers used in our simulation.

(a) The sequence alignment between PSM-β1 and PSM-β2. A ‘*’ denotes positions with identical residues, ‘:’ indicates strongly similar properties, and ‘.’ denotes weakly similar properties. Residue colors depict characteristics: black for hydrophobic, red for negatively charged, blue for positively charged, and green for polar. (b) The initial structures of PSM-β1 and PSM-β2 monomers used in our simulation. PSM-β2’s structure was retrieved from PDB:5KGZ, while PSM-β1’s was generated via AlphaFold prediction. The N-terminal Cα atom is highlighted by bead, with side-chains represented as sticks colored according to residue type: white for hydrophobic, red for negatively charged, blue for positively charged, and green for polar residues.

Table 1.

Details of each molecular system simulation, including the number of peptides in the system (System), the type of peptide (Peptide), dimensions of the cubic simulation box (Box Size), the number of independent DMD simulations performed for each system (DMD Runs), the duration of each independent DMD simulation (Time), and the cumulative total simulation time (Total Time).

System	Peptide	Box Size (nm)	DMD Run	Time (μs)	Total Time (μs)
1-peptide	PSM-β1	6.5	50	0.6	30
	PSM-β2	6.5	50	0.6	30
2-peptides	PSM-β1	7.5	50	1.0	50
	PSM-β2	7.5	50	1.0	50

2.2. DMD simulations

All simulations were performed at 300 K using the discrete molecular dynamics (DMD) algorithm[42, 46] along with the atomistic Medusa force field[47, 48]. Both DMD and traditional molecular dynamics (MD) are based on the same underlying physical principles, but they differ significantly in their approach to interatomic interactions[46]. In conventional MD, continuous potential functions are utilized, whereas DMD adopts step functions to represent these interactions. The step function potentials are derived from the Medusa force field[47, 48]. In this force field, bonded interactions, including bonds, bond angles, and dihedrals, are modeled using infinite square wells. Covalent bonds and bond angles typically have a single well, while dihedrals can feature multiple wells corresponding to cis or trans conformations. Nonbonded interactions, such as van der Waals forces, solvation, hydrogen bonding, and electrostatic terms, are represented by a series of discrete energetic steps that decrease in magnitude as the distance increases, reaching zero at the cutoff distance. The DMD framework operates through a series of collision events, during which two atoms encounter one another at a defined energy threshold and adjust their velocities according to the laws of conservation[49, 50]. By focusing on the two colliding atoms, forecasting their new interactions with adjacent atoms, and employing quick sort algorithms to identify subsequent collisions, DMD markedly enhances sampling efficiency. This method circumvents the need for frequent force and acceleration calculations, which typically occur every ~1–2 fs in standard MD simulations.

The Medusa force field has been validated for accurately predicting changes in protein stability due to mutations and assessing protein-ligand binding affinities[47, 48]. We employed the EEF1 implicit solvent model[51] for solvation and used a reaction-like algorithm to explicitly model hydrogen bond formation[42]. Electrostatic interactions were calculated using the Debye–Hückel approximation, with a Debye length of approximately 10 Å under physiological conditions. The DMD software is available to academic researchers at Molecules In Action (www.moleculesinaction.com). In our united-atom model with implicit water, the units of mass, time, length, and energy were set to 1 Da, ~50 fs, 1 Å, and 1 kcal/mol, respectively. The Medusa force field combined with the EEF1 implicit solvent model has shown reliable performance, accurately predicting protein folding[42], generating conformational ensembles consistent with experimental data[52, 53], and replicating variations in amyloid formation and secondary structures of calcitonin analogs[52, 54, 55] and amylin peptides[15, 56]. This approach’s accuracy was further confirmed by comparing DMD simulations with standard MD simulations of functional amyloid suckerin peptides[57] and pathological amylin peptides[56]. DMD’s speed and enhanced sampling efficiency have made it popular among our research team[9, 55, 58] and the broader scientific community[44, 59, 60] for studying protein folding and aggregation.

2.3. Analysis methods

The analysis of secondary structures was conducted using the DSSP (Dictionary of Secondary Structure of Proteins) method[61]. Hydrogen bonds were defined based on an N···O distance of less than 3.5 Å and an N–H···O angle exceeding 150°. Residue contacts were identified when the distance between heavy atoms of non-adjacent side chains or main chains was within 0.65 nm. Conformational cluster analysis for both PSM-β1 and PSM-β2 monomers was performed using the Daura algorithm [62], with a backbone deviation cutoff of 0.55 nm. By calculating the RMSD between different conformations, the algorithm identified their spatial similarities. Structures with a backbone RMSD smaller than 0.55 nm were considered similar and grouped together, while those with higher RMSD values were assigned to different clusters. Additionally, a two-dimensional (2D) free energy surface, known as the potential mean force (PMF), was created using the equation −RT ln P(x, y), where P(x, y) denotes the probability of observing a conformation with specific parameter values x and y.

3. Results and discussions

3.1. PSM-β1 monomer frequently switched between helix and β-sheet, while PSM-β2 monomer remained mostly helical.

The conformational dynamics of PSM-β1 and PSM-β2 monomers were investigated through fifty independent 600 ns DMD simulations to ensure thorough sampling. Equilibrium was assessed by tracking parameters such as the radius of gyration (Rg), backbone hydrogen bonds, heavy atom contacts, and secondary structure content (Figures S1–S2). Analysis of three randomly selected trajectories from these simulations showed that both PSM-β1 and PSM-β2 monomers reached equilibrium within the final 300 ns. Consequently, this 300 ns data was used for detailed conformational analysis. The significant conformational changes observed in each independent trajectory indicated that conformational bias from the initial states was reasonably avoided.

Time evolution of the secondary structures at the residue level revealed that PSM-β1 exhibited significant structural flexibility, frequently switching between α-helical and β-sheet conformations, with a slight preference for β-sheet stability (Figure 2a). In contrast, PSM-β2 predominantly maintained a stable α-helical conformation, especially in the N-terminal region (residues 1–20). However, the C-terminal region (residues 26–44) of PSM-β2 displayed more structural variability, transitioning between helical and β-sheet conformations (Figure 2b). The average secondary structure content of PSM-β1 and PSM-β2 monomers, calculated from the final 300 ns of simulation data, revealed that PSM-β1 primarily adopted unstructured conformations (~43.4%), with additional contributions from β-sheets (~23.9%) and helices (~20.9%) (Figure 2c). In contrast, PSM-β2 showed a higher preference for helical conformations (~43.0%), while unstructured conformations and β-sheets accounted for ~36.9% and ~10.4%, respectively (Figure 2c). The distribution of helix and β-sheet content indicated that PSM-β1 adopted a mixture of both secondary structures, whereas PSM-β2 was predominantly helical (Figure 2d). Interestingly, prior CD spectra showed that incubating both peptides with 50% TFE significantly enhanced their helical signals, indicating that most residues primarily adopted dynamic helical conformations[26, 38]. Without TFE, both PSM-β1 and PSM-β2 displayed a considerable amount of unstructured content; however, PSM-β2 exhibited stronger helical signals than PSM-β1, while PSM-β1 showed slightly higher β-sheet content than PSM-β2[26, 38]. These conformational characteristics aligned well with our simulation results.

Figure 2. Secondary structure analysis of PSM-β1 and PSM-β2 monomers during simulations.

(a-b) Time evolution of secondary structures for each residue in PSM-β1 and PSM-β2 monomers, with dynamic structures displayed at 200 ns intervals. Two trajectories were randomly selected from fifty independent discrete molecular dynamics (DMD) simulations for each molecular system. (c) Average secondary structure content for PSM-β1 and PSM-β2 monomers in saturated states. (d) Probability distribution functions (PDFs) illustrating the content of helical and β-sheet structures for each monomer, derived from the last 300 ns of fifty independent 600 ns DMD simulations. (e) Average propensity of each residue to adopt helical, β-sheet, and unstructured conformations. (f-g) Central structures and corresponding probabilities of the six most populated conformations for PSM-β1 (red) and PSM-β2 (blue) monomers. For clarity, the first Cα atom of PSM-β1 and PSM-β2 is highlighted with a bead.

Residue-specific secondary structure propensity analysis demonstrated that residues 1–35 of PSM-β1 were mainly unstructured, with propensities exceeding 40% (Figure 2e). Helices were concentrated in residues 3–10, 14–19, and 33–42, with propensities of ~20%-48%. In contrast, β-sheet structures were primarily observed in residues 25–32 and 36–43, with propensities ranging from ~25%-53%, while residues 5–18 displayed a modest β-sheet tendency (~15%-30%). Despite the high sequence homology between PSM-β1 and PSM-β2 (~72.7%), PSM-β2 showed distinct conformational behavior. The N-terminal region of PSM-β2, especially residues 5–17, exhibited a strong helical propensity, exceeding 80%. The C-terminal region of PSM-β2 had similar helix and β-sheet regions to PSM-β1, but PSM-β2 was more helical, while PSM-β1 showed a higher tendency towards β-sheet formation in the corresponding regions (Figure 2e).

The six most abundant conformations of PSM-β1 accounted for 35.6% of the total structural ensemble, with a notable presence of both helical and β-sheet structures (Figure 2f). In contrast, the six most populated PSM-β2 conformations made up 74.5% of the total, predominantly adopting helical arrangements. The conformations of PSM-β2, especially in the N-terminal regions, predominantly displayed helical structures consistent with NMR-determined results under the current TFE conditions[36]. Variations in the C-terminus conformation from our simulations may be due to TFE-induced secondary structure changes. In contrast, the monomer structure of PSM-β1 remains elusive, likely due to its dynamic conformational behavior observed in our simulations. Overall, the folding dynamics and structural analysis indicate that PSM-β1 monomers are highly dynamic, frequently switching between helix and β-sheet structures, while PSM-β2 monomers tend to remain primarily in helical conformations.

3.2. Hydrophobic residue interactions facilitated the formation of a β-hairpin motif in PSM-β1 and PSM-β2 monomers, promoting the helix-to-β-sheet conversion around the PSM-β1 N-terminus.

To further elucidate the key interactions governing the dynamic structures of PSM-β1 and PSM-β2 monomers, we performed a residue-pairwise contact frequency analysis for both systems (Figure 3). Both PSM-β1 and PSM-β2 exhibited similar helical contact patterns along the diagonal, particularly in regions encompassing residues 4–23 and 31–42 (conformations 1 and 2 in Figure 3a&b). Notably, the helical contact frequency was more pronounced in PSM-β2 for the former region, consistent with the secondary structure analysis. These helical motifs displayed an amphipathic character, with one face enriched in hydrophobic residues, while the opposing face was dominated by charged and polar residues.

Figure 3. Residue-pairwise contact frequency analysis of PSM-β1 and PSM-β2 monomers.

(a-b) The frequency of residue-pairwise contacts within PSM-β1 and PSM-β2 monomers, computed from the last 300 ns of 50 independent 600 ns discrete molecular dynamics (DMD) simulations after reaching equilibrium. The lower diagonal represents main-chain contacts, while the upper diagonal reflects side-chain contacts. High-contact-frequency regions corresponding to specific structured motifs are labeled on the contact maps, with representative secondary structures and corresponding amino acid sequences presented. To enhance clarity, residues in the structured regions are color-coded by type: hydrophobic (black), polar (green), positively charged (blue), and negatively charged (red). The first Cα atom is highlighted as a bead. PSM-β1 and PSM-β2 monomers are colored red and blue, respectively.

Due to the high sequence homology in the C-terminal regions of PSM-β1 and PSM-β2, both monomers displayed a prominent β-hairpin contact pattern, oriented perpendicular to the diagonal, involving β-strands formed by residues 26–33 and 36–43 (conformation 3 in Figure 3a&b). PSM-β1, however, displayed stronger contact frequencies within this β-hairpin motif compared to PSM-β2, suggesting a more stable and favorable β-hairpin conformation in PSM-β1. Additionally, PSM-β1 featured another prominent β-hairpin structure involving β-strands formed by residues 6–17 and 22–33 (conformation 4 in Figure 3a). In contrast, the PSM-β2 monomer rarely adopted β-sheet formations in this region, as residues 1–25 predominantly maintained a helical conformation, with only a faint β-hairpin pattern observed between residues 21–24 and 27–30 (conformation 4 in Figure 3b).

Beyond these β-hairpin formations, intra-peptide β-sheets were observed in PSM-β1, involving contacts between residues 7–13 and 37–43, as well as residues 3–6 and 40–43 (conformations 5 and 6 in Figure 3a). These intra-peptide β-sheets were not observed in PSM-β2. Analysis of residue-pairwise contact motifs indicated that the β-sheet formations were primarily stabilized by hydrophobic interactions, with some contribution from partial charge interactions. Overall, our residue-pairwise interaction analysis revealed that the conserved C-terminal residues 26–44 of both PSM-β1 and PSM-β2 could adopt dynamic β-hairpin conformations. In PSM-β1, the formation of the C-terminal β-hairpin appeared to facilitate helix-to-β-sheet transitions in the N-terminal region, a process that was not widely observed in PSM-β2.

3.3. Dimerization strengthened the stability of β-sheets in PSM-β1 and PSM-β2 by forming inter-molecular contacts and backbone hydrogen bonds.

To explore the conformational dynamics during the aggregation of PSM-β1 and PSM-β2, their homodimerization was investigated using fifty independent DMD simulations, each lasting 1000 ns. The time evolution of structural parameters—such as the radius of gyration, the total number of backbone hydrogen bonds, heavy atom contacts, and secondary structure content—was analyzed for three trajectories randomly selected from the simulation pool. The absence of significant shifts in these parameters during the last 500 ns indicated that equilibrium had been reached (Figures S3–S4). Thus, the final 500 ns of each DMD trajectory was used for the conformational analysis of PSM-β1 and PSM-β2 dimers.

The dimerization dynamics of PSM-β1 and PSM-β2 were investigated by monitoring the time evolution of secondary structure, as well as the total number of backbone hydrogen bonds and heavy atom contacts formed both intra- and inter-molecularly (Figure 4a&b). Both peptides demonstrated prion-like aggregation behavior, where two isolated PSM-β1 and PSM-β2 molecules rapidly assembled into stable oligomers, characterized by the formation of numerous inter-peptide backbone hydrogen bonds and heavy atom contacts. In the early stages of aggregation, inter-peptide heavy atom contacts formed first, followed by the establishment of inter-peptide backbone hydrogen bonds and β-sheet structures. This progression suggested that aggregation promoted a conformational shift toward β-sheet formation. Unlike the isolated monomers, which exhibited frequent dynamic structural changes in their β-sheet regions (Figure 2a&b), the β-sheet conformations of PSM-β1 and PSM-β2 in the dimeric state were more stable and exhibited fewer alterations upon reaching equilibrium, indicating that dimerization enhanced β-sheet stability.

Figure 4. Homo-dimerization dynamics analysis for PSM-β1 and PSM-β2.

(a-b) The dimerization dynamics and conformational changes of PSM-β1 and PSM-β2 were monitored by tracking the time evolution of secondary structures for each residue (first row), in addition to measuring the total number of intra- and inter-peptide hydrogen bonds and contacts (second row). Snapshots at 250, 500, 750, and 1000 ns are provided to illustrate the aggregation process (third row). The N-terminal Cα atom of each peptide is highlighted by a bead. (c-d) The free energy landscape of dimerization for PSM-β1 (left) and PSM-β2 (right) is shown as a function of the number of inter-peptide contacts against the average ratios of β-sheet (upper) and helix (bottom) content. Four representative snapshots corresponding to distinct minimal energy regions, as labeled on the free energy surface, are displayed on the right.

To further elucidate the dimerization process, free energy landscapes for both PSM-β1 and PSM-β2 dimers were constructed by plotting inter-peptide contacts against the average helix content and β-sheet ratio, using data from all 1000 ns of fifty independent DMD trajectories (Figure 4c&d). Conformations with fewer inter-peptide contacts (~0–30) were associated with high free energy, indicating that monomeric states were thermodynamically unfavorable. As the number of inter-peptide contacts increased, the most populated low-energy conformations displayed a transition from helical to β-sheet structures. The lowest-energy basin for PSM-β1 was predominantly enriched in β-sheet content, while PSM-β2 exhibited a higher proportion of helical structures. In summary, the dimerization dynamics and free energy landscape analysis suggested that both PSM-β1 and PSM-β2 were prion-like amyloid peptides that readily aggregated into β-sheet-rich structures, albeit with distinct structural preferences.

3.4. Structural reorganization of PSM-β1 and PSM-β2 upon homo-dimerization enhanced β-sheet formation and reduced unstructured conformations.

The secondary structure analysis demonstrated that homo-dimerization notably altered the conformational landscapes of PSM-β1 and PSM-β2 by increasing β-sheet content and reducing unstructured regions, while minimally affecting helical content (Figure 5a&b). Dimeric PSM-β1 and PSM-β2 showed enhanced β-sheet content, reaching 31.3% and 19.0%, respectively, compared to 23.9% and 10.4% in their monomeric forms. At the residue level, dimerization in PSM-β1 increased β-sheet propensity particularly in residues 11–18 and 24–30, while PSM-β2 displayed increased β-sheet formation in residues 20–34, notably residues 26–31, which increased by over 20%. Concurrently, unstructured conformations decreased in dimeric forms, with PSM-β1 and PSM-β2 showing reductions to 34.6% and 31.2%, respectively, from 43.4% and 36.9% in the monomeric states. Notable declines in unstructured propensity were observed in residues 4–18 and 23–30 of PSM-β1 and residues 24–32 of PSM-β2. Dimerization of PSM-β1 and PSM-β2 led to a conformational conversion toward β-sheets, which aligned with previous experimental observations of increased β-sheet content during the aggregation of both peptides, as indicated by CD and FTIR measurements[25, 37].

Figure 5. Conformational analysis of monomeric and dimeric PSM-β1 and PSM-β2.

(a) The average content of each secondary structure for monomeric and dimeric PSM-β1 and PSM-β2. (b) The average propensity of each residue in PSM-β1 and PSM-β2 to adopt helix (top), β-sheet (middle), and unstructured (bottom) conformations in both monomeric and dimeric states. (c-d) The averaged solvent accessible surface area (SASA) and the average number of intra- and inter-peptide contacts for each residue in PSM-β1 (top) and PSM-β2 (bottom) across monomeric and dimeric states. For each molecular system, only the final 300 ns of simulation data from 50 independent 600 ns DMD trajectories are utilized for monomer analysis, while the last 500 ns of simulation data from 50 independent 1000 ns DMD trajectories are employed for dimer analysis.

Helical content showed only minor changes; PSM-β1’s helical content slightly increased from 20.9% to 23.6% upon dimerization, while PSM-β2 exhibited a small reduction from 43.0% to 41.6%. Prior experimental measurements revealed that PSM-β2 aggregates exhibited greater helical content than those of PSM-β1, consistent with our observation that PSM-β2 dimers retained more helical structure than PSM-β1 dimers[25, 37]. Detailed residue-level analysis further indicated a ~15.8% increase in helical propensity for residues 35–43 in PSM-β1, accompanied by a ~6.0% decrease for residues 13–20, while PSM-β2 displayed a ~9.0% reduction around residues 17–23. Secondary structure differences between PSM-β1 and PSM-β2 were most apparent in residues 1–20, with residues 21–44 displaying similar structural trends with minor variations in propensity. Residues 1–20 in PSM-β2 predominantly adopted a helical structure in both monomeric and dimeric states, while PSM-β1 exhibited a mix of helices and β-sheets in this region. Both PSM-β1 and PSM-β2 dimers displayed prominent β-sheet formation around residues 21–33, a region largely unstructured in their monomers, and residues 34–44 showed a preference for both helical and β-sheet formations, with helices being more prevalent.

Analysis of solvent-accessible surface area (SASA) indicated that the most populated β-sheet region in residues 21–33 of both peptides exhibited reduced surface exposure in the dimer compared to the monomer, along with an increased number of inter-peptide contacts, suggesting a critical role for this region in facilitating dimerization (Figure 5c&d). In contrast, the N-terminal residues 1–20 showed more intra-peptide than inter-peptide contacts in both peptides, implying a limited role in promoting self-assembly. The structural similarities between PSM-β1 and PSM-β2 from residues 21–44, but not in the N-terminal 1–20 region, likely reflected sequence conservation at the C-terminus, contrasting with sequence divergence in the N-terminal region.

3.5. The conserved C-terminus facilitated the self-assembly of both PSM-β1 and PSM-β2 into β-sheet structures, while the N-terminus participated in β-sheet formation for PSM-β1 but remained primarily helical in PSM-β2.

To elucidate the structural determinants underlying PSM-β1 and PSM-β2 dimer stabilization, we analyzed intra- and inter-peptide residue-pairwise heavy atom contact frequencies in their homodimers (Figure 6). For PSM-β1 dimers, intra-peptide contacts revealed two prominent helical interaction regions between residues 2–23 and 33–42 (conformations 1 and 2 in Figure 6a) alongside a notable β-hairpin structure formed by interactions between residues 4–21 and 23–40 (conformation 5 in Figure 6a). Additionally, two moderately populated β-hairpin motifs were observed in the terminal regions (residues 24–30 vs. 37–43 and 2–9 vs. 12–19), as well as a weakly populated parallel β-sheet between residues 4–12 and 34–42 (conformations 3, 4, and 6 in Figure 6a). In the inter-peptide contact map, β-sheet formations predominantly involved residues 21–30, showing significant frequencies for anti-parallel β-sheets between residues 25–30 and 25–30 and a short parallel β-sheet spanning residues 5–8 and 39–42 (conformations 10 and 11 in Figure 6a). Additional inter-peptide contacts included less frequent parallel β-sheets (residues 3–13 vs. 26–36 and 25–32 vs. 25–32, conformations 7 and 8 in Figure 6a) and anti-parallel β-sheets (residues 26–33 vs. 25–42, 13–18 vs. 26–31, and 6–17 vs. 26–37, conformations 9, 12, and 13 in Figure 6a). Notably, residues 21–33 displayed high frequency of both intra- and inter-peptide contacts within β-sheet formations, underscoring the critical role of this segment in facilitating PSM-β1 dimerization and subsequent β-sheet stabilization.

Figure 6. Residue-pairwise contact frequency analysis within PSM- β1 and PSM- β2 homodimers.

(a-b) The frequency of residue-pairwise contacts formed by atoms from the intra-chain (lower diagonal) and inter-chain (upper diagonal) within the PSM- β1 and PSM- β2 homodimers are computed using the last 500 ns simulation data from 50 independent 1000 ns DMD trajectories after reaching a steady state. Representative structured motifs and corresponding amino acid sequences, exhibiting high-contact-frequency patterns labeled on the surface of the contact map, are also presented. To enhance clarity, residues in the structured regions are color-coded by type: hydrophobic (black), polar (green), positively charged (blue), and negatively charged (red). The first Cα atom is highlighted as a bead. PSM-β1 and PSM-β2 dimers are colored red and blue, respectively.

The PSM-β2 dimer displayed two prominent helical contact regions around residues 2–23 and 33–42, with higher contact frequency compared to PSM-β1 (conformations 1 and 2 in Figure 6b). Intra-peptide contacts indicated that β-sheet formations within PSM-β2 were limited to two relatively weak β-hairpin motifs, observed between residues 24–32 and 35–43, and between 16–23 and 26–33 (conformations 3 and 4 in Figure 6b). The N-terminal residues 1–20 predominantly maintained helical conformations, while inter-peptide β-sheets formed between residues 21–44. These β-sheets included anti-parallel arrangements, such as between residues 23–32 and 28–37, 25–31 and 28–33, and 19–25 and 26–32 (conformations 6–8 in Figure 6b), as well as a parallel β-sheet between residues 38–43 and 38–34 (conformation 5 in Figure 6b). Similar to PSM-β1, the region spanning residues 21–33 was the most frequently involved in β-sheet formation within the PSM-β2 dimer, while the N-terminal region remained uninvolved in β-sheet formation, showing a strong preference for helical conformations.

Our residue-pairwise contact frequency analysis revealed that dimerization in both PSM-β1 and PSM-β2 promoted β-sheet conformations in the conserved C-terminal residues (21–44), especially within residues 21–33. However, the self-aggregation propensity of the N-terminal residues (1–20) in both peptides was relatively weak. In PSM-β1, the N-terminus formed β-sheets only when interacting with the C-terminal region, while the N-terminus of PSM-β2 did not participate in β-sheet formation. This marked difference in aggregation potential between the N- and C-termini likely contributed to their distinct roles in S. aureus biofilm dynamics: the C-terminal regions promoted biofilm formation, while the N-terminal regions regulated biofilm stability. Given its higher aggregation tendency, PSM-β1 was likely more effective in promoting biofilm growth, while the helical nature of PSM-β2’s N-terminus may have enhanced its capacity for biofilm dispersal. This hypothesis aligned with experimental findings showing that PSM-β1 more effectively promoted biofilm formation at moderate concentrations, whereas PSM-β2 demonstrated greater potency in biofilm dispersal at higher concentrations[33, 34]. Interestingly, prior experimental studies on the amyloid aggregation of PSM-β1 and PSM-β2 revealed that PSM-β1 exhibited a stronger tendency to form β-sheets, whereas PSM-β2 displayed a greater propensity to maintain a helical structure[25, 36, 39], consistent with our simulation results. Furthermore, chemical kinetics analysis of the seeding mechanism highlighted distinct fibrillization pathways: PSM-β1 fibrils catalyzed aggregation via secondary nucleation and fibrillar elongation, while PSM-β2 fibrils predominantly facilitated fibrillization through fibrillar elongation alone[25]. These differences could be attributed to the distinct aggregation characteristics of the N-terminal domains in PSM-β1 and PSM-β2. Specifically, the N-terminal residues of PSM-β1 participated in β-sheet formation by interacting with the amyloidogenic C-terminal domain, whereas the N-terminal residues of PSM-β2 largely retained a helical conformation and did not contribute to β-sheet aggregation. Although the fibrillar structures of both peptides have yet to be determined experimentally, the presence of the N-terminus in PSM-β1 fibrils may shield the lateral surface, thereby reducing seeding capability when the C-terminus is buried. Overall, our findings suggested that the regulatory roles of PSM-β1 and PSM-β2 in biofilm dynamics were driven by their distinct aggregation propensities in the N- and C-terminal regions.

The conserved C-terminal residues 21–33, acting as the amyloidogenic core, drove the aggregation of PSM-β1 and PSM-β2, exhibited a strong intermolecular tendency to form β-sheet structures, and significantly interacted with other regions to further promote β-sheet formation. Although the critical role of this region in driving β-sheet aggregation was primarily investigated through PSM-β1 and PSM-β2 monomers and dimers, it was expected that the recruitment of additional peptides could also convert into β-sheet formations and grow into larger aggregates. For example, the amyloidogenic core region observed in the dimerization of Aβ, and hIAPP was similarly crucial for β-sheet formation in larger aggregates during early nucleation stages[10, 15]. Furthermore, numerous studies have shown that the amyloidogenic core regions of amyloid peptides, both in oligomeric and fibrillar states, can recruit isolated peptides to extend along the β-sheet edges with limited sensitivity to the amino acid sequence[58, 63, 64]. Capping or binding of the amyloidogenic core regions of Aβ and hIAPP by amyloid-resistant peptides, small molecules, and nanoparticles could effectively prevent their abnormal pathological aggregation[65–69]. Therefore, targeting the conserved C-terminal residues 21–33 of both PSM-β1 and PSM-β2 with amyloid inhibitors (e.g., amyloid-resistant peptides, small molecules, nanoparticles) could offer a promising therapeutic strategy for combating biofilm-associated pathogenicity. This could be achieved by capping their growth edges with amyloid-resistant peptides, shielding their exposed surface area with small molecules, or disrupting their β-sheet formations using nanoparticles[70], thereby preventing the recruitment of additional peptides and the formation of larger aggregates.

4. Conclusion

In conclusion, our comprehensive atomistic DMD simulations provide critical insights into the conformational dynamics and aggregation mechanisms of PSM-β1 and PSM-β2, elucidating their key roles in the biofilm formation of S. aureus[26, 29]. These simulations revealed distinct structural behaviors of each peptide, with PSM-β1 frequently transitioning between helical and β-sheet conformations, showing a stronger propensity for β-hairpin formation. This conformational flexibility facilitated its dimerization and enhanced β-sheet stability across the peptide, especially within residues 11–18 and 21–33. In contrast, PSM-β2 maintained a predominantly stable helical conformation, particularly around N-terminal residues 1–20, which contributed minimally to β-sheet formation. However, its conserved C-terminus robustly supported dimerization, leading to extensive inter-peptide hydrogen bonding that reinforced β-sheet structures around the C-terminal region. A notable finding was the critical role of residues 21–33 in both peptides, which underwent a transition from unstructured conformations in monomers to β-sheets upon dimerization. This region not only drove self-aggregation into β-sheets but also stabilized β-sheet formation in adjacent regions. PSM-β1 demonstrated a stronger tendency for aggregation, with β-sheet formation extending throughout the peptide, while PSM-β2 primarily contributed to β-sheet structures via its C-terminus, retaining a helical N-terminus. These differences highlight the unique conformational dynamics of each peptide and underscore their respective contributions to the pathogenic biofilm architecture of S. aureus. Our findings suggest that effective therapeutic strategies targeting fibrous aggregation in PSM-β1 and PSM-β2 should focus on residues 21–33. By clarifying the molecular basis of their conformational flexibility and aggregation behavior, this study provides a foundational understanding that can inform drug design aimed at disrupting biofilm formation, addressing a critical challenge in treating S. aureus infections. Ultimately, this work advances the molecular understanding of biofilm-associated pathogenicity and opens new avenues for targeted interventions against biofilm-related diseases.

Supplementary Material

Appendix A. Supplementary data

Supplementary figures (PDF)

Data availability

Data will be made available on request.