Surface Chirality Enables MtlA‐Dependent Antibacterial Activity and Immune Remodeling by Gold Nanoparticles

ABSTRACT

Skin wound infections persist through a self‐reinforcing “infection‐inflammation” cycle that compromises tissue repair. Here, we demonstrate that surface chirality of nanomaterials serves as an independent parameter to disrupt this loop through targeted metabolic interference. Using gold nanoparticles (GNPs) as a model, we show that D‐chiral GNPs (GNP‐D) exhibit superior efficacy over their L‐enantiomers (GNP‐L) in repressing bacterial basal metabolism and impeding near‐surface motility. Integration of transcriptomic profiling and molecular dynamics identifies the mannitol‐specific phosphotransferase transporter MtlA as a pivotal target, where GNP‐D induces enhanced structural perturbations at functional residues compared to GNP‐L. In a murine infection model, GNP‐D significantly reduces bacterial burden and accelerates wound closure while repolarizing the wound microenvironment from a persistent pro‐inflammatory state toward a pro‐resolving and regenerative phenotype. Notably, genetic ablation of mtlA abrogates the chiral bias across all metabolic and immune readouts, providing evidence that this advantage is mediated via the MtlA/PTS axis. These results establish a “chirality‐membrane‐protein coupling” framework for the rational design of targeted antimicrobial and immunomodulatory nanomedicines.

1. Introduction

Cutaneous and wound infections often enter a vicious “infection‐inflammation” cycle [1]. Sustained accumulation of pro‐inflammatory signals hinders granulation and re‐epithelialization, predisposing wounds to chronicity and secondary colonization. Among causative bacteria, Staphylococcus aureus (S. aureus) is a leading pathogen owing to strong adhesion and a marked propensity for biofilm formation [2]. This adhesion‐biofilm phenotype promotes persistent colonization and treatment failure; consequently, methicillin‐resistant S. aureus (MRSA) is particularly difficult to eradicate and shows high recurrence rates [3]. Biofilms are detected in ∼60% of chronic wounds but in only ∼10% of acute wounds, and the lack of specific clinical signs likely leads to underestimation [4]. Even with standard care that includes systemic or topical antibiotics, antimicrobial dressings and debridement, clinical benefit is often limited by biofilm barriers and inadequate drug penetration, leading to variable efficacy and frequent relapse [5]. Beyond local failure, antibiotic‐resistant pathogens persist, disseminate into deeper wound niches, and evade immune clearance, thereby amplifying inflammation and delaying repair; in severe cases, they cause death. Globally, drug‐resistant infections were directly responsible for an estimated 1.27 million deaths in 2019, with projections exceeding 10 million annually by 2050 [6]. These outcomes underscore the need to devise strategies that cross biofilm barriers, engage key targets, and improve the repair microenvironment.

Nanomaterials offer a materials‐based approach to these problems. Nanoparticles (NPs) can be tuned in size, shape, and surface chemistry. Such tunability can enhance lesion localization and directly perturb bacterial membranes and biofilms [7]. Recent studies revealed that surface topology of nanomaterials is becoming a crucial matter, since the geometry and ligand coordination can control the strategy and strength of binding between NPs and biological interfaces [8]. In mammalian systems, nanoparticles rapidly adsorb proteins from biological fluids to form a protein corona, which reshapes cellular recognition, uptake, metabolism, and immune responses [9]. By contrast, bacterial envelopes lack such a dynamic corona but present densely packed lipids, peptidoglycan, and membrane transporters that gate metabolic flux and stress responses [10]. Among these, the mannitol phosphotransferase system (PTS) transporter MtlA channels carbon and phosphate flux, linking metabolism to envelope homeostasis [11]. At this interface, membrane proteins can act as critical intermediaries that modulate NP recognition, binding, and downstream functional consequences. Consequently, in bacteria, NPs act chiefly by damaging membranes and ionic balance, remodeling metabolism, and weakening biofilm [12]. Given the close interplay between bacterial persistence, biofilm formation, and host inflammatory tone, perturbations of bacterial metabolism and envelope integrity have the potential to extend beyond antibacterial effects and influence the wound immune microenvironment.

Against this background, chirality can be used as an independent and tunable parameter at the nano‐bio interface [13]. Notably, advances in observational and analytical techniques have brought the relationship between protein three‐dimensional structure and functional stability to the fore, establishing a new paradigm for studying physiological change. When other physicochemical variables are held constant, altering only the enantiomeric configuration (L or D) can elicit biological differences [14]. Studies have reported that chiral NPs act differently at multiple biological levels, but most studies are confined to a single biological scale or experimental context [15]. Prior work shows that chiral nanoparticles can alter membrane binding, protein conformations, and immune readouts, yet most studies isolate a single scale (for example, in vitro only or a single assay) and leave target specificity unresolved [16]. Therefore, systematic and cross‐scale evidence remains limited.

Based on the progress of recent studies, we posit a “chirality‐membrane‐protein coupling” hypothesis: with particle size, shape, surface potential, and ligand chemistry held constant, only topological chirality varies to test enantiomer specific interactions at S. aureus membrane‐protein microdomains. Using gold nanoparticles (GNPs) as a model, we link in vitro and in vivo readouts across physicochemical checks and antimicrobial assays (minimum inhibitory concentrations, MIC; colony‐forming units, CFU) in S. aureus and MRSA. We further integrate morphology by scanning/transmission electron microscopy (SEM/TEM), biofilm and protein‐leakage measurements, adenosine triphosphate (ATP) and reactive oxygen species (ROS) quantification, and near‐surface tracking by total internal reflection fluorescence (TIRF) microscopy. D‐chiral GNPs (GNP‐D) exhibit stronger antibacterial and antibiofilm activity than L‐chiral GNPs (GNP‐L). RNA‐seq implicates the mannitol PTS transporter MtlA as a candidate node linking chirality to bacterial metabolism. Molecular dynamics (MD) reveals energetic/geometric advantages for the D form with changes in secondary structure, residue fluctuations, and the local environment of Cys425 and His487. In a murine wound model, GNP‐D reduce bacterial burden and accelerate healing with acceptable safety; day‐5 flow cytometry indicates early clearance with dampened inflammation, and day‐9 immunohistochemistry shows progression to resolution/remodeling. The D/L advantage disappears in the RH (mtlA‐deficient) background, indicating target specificity and supporting chirality as a functionally independent design variable under otherwise controlled physicochemical conditions. Thus, we propose that GNPs interact with S. aureus membrane proteins, with the MtlA/PTS axis serving as a key mechanistic node for the observed chirality‐dependent antibacterial effects (Scheme 1). Overall, our work builds a cross‐scale evidence chain for chirality and establishes chirality as an independent design variable for antibacterial nanomaterials, opening routes to antimicrobial and anti‐inflammatory therapies that complement antibiotics.

SCHEME 1.

Dual Antibacterial and Anti‐inflammatory Mechanism of Chiral Gold Nanoparticles. Dual Antibacterial and Anti‐inflammatory Mechanism of Chiral Gold Nanoparticles (GNP‐L/D). Right: Phase‐dependent Immune Remodeling in the Infected Wound Microenvironment Induced by Chiral GNPs. Early Phase (Day 5): Enhanced Bacterial and Biofilm Damage Promotes Antigen Release and Presentation, Accompanied by Increased MHC‐II (I‐A/I‐E) Expression, Restrained Pro‐inflammatory Signaling, and Reorganization of Innate and Adaptive Immune Compartments, Collectively Supporting Efficient Bacterial Clearance. Late Phase (Day 9): Inflammation Further Resolves, with a Shift toward Repair‐associated Immune States Characterized by Reduced CD86, Elevated CD206, and Improved Collagen Organization, Indicating Progression toward Tissue Remodeling and Wound Healing. Overall, GNP‐D Drive a More Coordinated Transition from Early Clearance to Late Repair than GNP‐L.

2. Results and Discussion

2.1. Preparation and Characterization of Chiral Gold Nanoparticles

To introduce “chirality” as an independent and controllable materials parameter, we first prepared cubic gold nanoparticles (cube‐GNPs) by seed‐mediated growth, as detailed in the Methods. Surface chirality was then induced with L‐ or D‐cysteine at four dosages (0.1, 1, 2.5, and 5 µL) to generate paired series (hereafter referred to as 0.1L, 1L, 2.5L, 5L and 0.1D, 1D, 2.5D, 5D, respectively) (Figure 1A) [17]. Physicochemical properties across the paired GNP‐L and GNP‐D groups were systematically characterized to minimize potential confounding factors and to support surface chirality as the primary design variable. Dynamic light scattering (DLS) analysis showed that the hydrodynamic diameters of GNP‐D and GNP‐L were 287.0 ± 0.8 nm and 275.6 ± 0.7 nm, respectively, with both exhibiting narrow polydispersity indices (PDI = 0.14 ± 0.02 and 0.10 ± 0.03 for GNP‐D and GNP‐L, respectively; Figure 1B). Correspondingly, zeta potential measurements yielded comparable surface charges of 39.8 ± 6.0 mV for GNP‐D and 37.7 ± 5.3 mV for GNP‐L (Figure 1C). These strongly positive potentials confirm successful cysteine functionalization and suggest robust colloidal stability. Furthermore, UV–vis spectra exhibited nearly identical surface plasmon resonance (SPR) profiles, with only minor, reproducible shifts attributed to ligand‐mediated dielectric effects rather than changes in particle volume (Figure 1D).

FIGURE 1.

Preparation and Physicochemical Characterization of Chiral Gold Nanoparticles. (A) Schematic Illustration of the Synthesis of GNP‐L and GNP‐D via Seed‐mediated Growth and Surface Chirality Induction. (B) Hydrodynamic Size Distributions of GNP‐D (287.0 ± 0.8 Nm, PDI = 0.14 ± 0.02) and GNP‐L (275.6 ± 0.7 Nm, PDI = 0.10 ± 0.03) Measured by DLS. (C) Zeta Potentials of GNP‐D (39.8 ± 6.0 mV) and GNP‐L (37.7 ± 5.3 mV) Measured by DLS. Data Are Presented as Mean ± SD (n = 3). (D) UV–vis Absorption Spectra of GNP‐L and GNP‐D. (E, F) Representative Low‐magnification (E) and High‐magnification (F) TEM Images of the Synthesized Nanoparticles. Scale Bars: 1 µm (E) and 200 Nm (F). (G, H) Circular Dichroism (CD) Spectra (G) and Dissymmetry Factor (G‐factor) Plots (H) of GNP‐L and GNP‐D.

Elemental composition and distribution were verified by SEM–EDS and elemental mapping, confirming gold (Au) as the primary metallic component for both enantiomers (Figure S1). Although the initial cubic seeds undergo ligand‐directed anisotropic growth to form the final branched/star‐like structures, high‐magnification TEM and SEM imaging (Figure 1E,F) revealed that the macroscopic core dimensions and branching geometries are indistinguishable between GNP‐L and GNP‐D. This confirms that the structural variations between the two enantiomers are confined to surface chiral arrangements. To quantitatively evaluate the optical activity, circular dichroism (CD) spectra and corresponding dissymmetry factors (g‐factors) were recorded (Figure 1G,H). GNP‐L and GNP‐D exhibited symmetrical mirror‐image CD responses across the visible‐to‐near‐infrared region, with comparable peak amplitudes and opposite signs. This chiroptical symmetry, coupled with the near‐identical hydrodynamic sizes (Figure 1B), zeta potentials (Figure 1C), and SPR profiles (Figure 1D), demonstrates that surface chirality has been successfully isolated as a single‐variable materials parameter. Together, these results indicate that GNP‐L and GNP‐D possess broadly comparable physicochemical profiles while displaying opposite chiroptical responses, thereby providing a reasonable basis for subsequent chirality‐dependent biological comparisons.

2.2. Chirality‐Dependent Antibacterial and Immunomodulatory Activities of GNPs

2.2.1. In Vitro Antibacterial Activity of GNP‐L and GNP‐D

Gold nanoparticles (GNPs) are widely reported to exhibit antibacterial activity [18, 19]. However, the antibacterial mechanisms associated with the physicochemical attribute of chirality remain poorly explored. We therefore selected the classical Gram‐positive pathogen S. aureus and MRSA as model pathogens. These pathogens are leading global causes of skin and soft tissue, wound, and device‐associated infections [20]. S. aureus combines genomic plasticity, rapid clonal virulence, biofilm persistence, and broad antibiotic resistance [2]. It shifts between commensal and invasive states through immune evasion and metabolic crosstalk with the host [21]. After establishing material stability and controlling other physicochemical variables, we further standardized concentration and the extent of chiral induction. To isolate concentration effects between groups, we performed MIC assays. In the OD₆₀₀ matrix, blue indicated lower bacterial density and stronger growth inhibition, whereas red indicated higher bacterial density and bacterial proliferation (Figure S2). Consistent with a chirality effect, GNP‐D generally required lower MICs than their GNP‐L counterparts. For example, in S. aureus, the MIC for 1D was approximately half that of 1L. Notably, while MRSA exhibited higher survival tenacity in the CFU assays (as shown in Figure 2B), it still followed the chirality‐sensitive pattern where GNP‐D outperformed GNP‐L in MIC tests. It is important to note that while OD ₆₀₀‐based measurements reflect the integration of metabolic activity and biomass growth, CFU counts provide a definitive assessment of cell viability and bactericidal efficiency [22]. These results defined the working concentration window for subsequent experiments and provided initial evidence that GNP‐D exhibits stronger antibacterial potency than GNP‐L. To assess concentration‐dependent cytotoxicity, we performed Cell Counting Kit‐8 (CCK‐8) assays. Across NIH‐3T3 (NIH/3T3 mouse embryonic fibroblasts) and L929 (NCTC clone 929 mouse fibroblasts), all groups maintained ≥80% viability, meeting the non‐cytotoxic criterion (Figure S3). Baseline antibacterial phenotypes were assessed by CFU counting (Figure 2A,B). At matched doses, both enantiomers effectively reduced the survival of S. aureus and MRSA, with 1D consistently demonstrating superior antibacterial efficacy over 1L. Under the optimized dosage (1 L/D), 1D achieved the highest antibacterial rate against S. aureus. As expected from precise single‐cell survival assays, the corresponding rate against MRSA was slightly lower (96.0 ± 0.8%), likely due to its intrinsic resistance mechanisms. Furthermore, it is noteworthy that the antibacterial potency did not follow a strictly linear pattern with respect to the chiral ligand induction dosage. Instead, a ‘hook effect’ was observed, wherein the 1 L/D groups exhibited the most pronounced antibacterial activity across all conditions, significantly outperforming the groups with excessive ligand densities (2.5 and 5 L/D). This ‘dosage paradox’ is primarily attributed to ligand‐induced self‐aggregation of the highly modified branched GNPs in the complex bacterial culture media, which reduces their effective interfacial contact with the bacteria [23]. Consequently, the 1 L/D dosage represents an optimized therapeutic window that balances chiral ligand density with colloidal stability, maximizing the targeted metabolic interference.

FIGURE 2.

Chirality‐dependent antibacterial and immunomodulatory activities of GNPs in vitro and in vivo. (A) Representative colony‐forming unit (CFU) plates of S. aureus treated with vehicle (CON) or chiral GNPs induced with increasing doses of L‐ or D‐cysteine (0.1L, 1L, 2.5L, 5L; 0.1D, 1D, 2.5D, 5D). (B) Corresponding CFU plates for MRSA under identical treatment conditions. Right panels in (A) and (B) show quantitative analysis of relative antibacterial rates. (C) Intracellular ATP levels of S. aureus and MRSA following treatment with GNP‐L or GNP‐D at different induction doses. (D) Intracellular ROS levels of S. aureus and MRSA under the same conditions as in (C). (E) Schematic of the RAW264.7 macrophage–bacteria co‐culture model used to assess macrophage inflammatory responses after GNP treatment. RAW264.7 cells were co‐incubated with S. aureus for 2 h, followed by GNP‐L or GNP‐D treatment and subsequent RT‐qPCR analysis. (F, G) Relative mRNA expression levels of Il1b and Il10 in RAW264.7 macrophages at 6 h (L6, D6) and 24 h (L24, D24) after treatment with GNP‐L or GNP‐D. Data are presented as fold changes relative to the untreated control group (set to 1.0) after normalization to Gapdh. All reactions were performed in triplicate and data are shown as mean ± SD. (H) Schematic of the murine full‐thickness dorsal wound infection model. Mice received S. aureus inoculation on day 0, local GNP‐L or GNP‐D treatment on days 1 and 3, and wound swab collection on day 9. Created in BioRender. (I) Representative CFU plates of wound swabs collected on day 9 from untreated (SA), GNP‐L–treated (1L), and GNP‐D–treated (1D) groups, with corresponding quantification of relative antibacterial rates. (J) Masson's trichrome staining of day‐9 wound sections from SA, 1L, and 1D. Collagen fibers are stained blue; muscle/cytoplasm red; nuclei dark. Asterisks mark collagen‐rich regions. Scale bar, 100 µm. (K) Schematic illustration summarizing the proposed antibacterial and immunomodulatory mechanisms of chiral GNPs. GNP‐D shows stronger interference with bacterial membrane and metabolic processes, leading to ATP depletion, ROS elevation, protein leakage, RNA damage, and biofilm inhibition, while also promoting balanced macrophage inflammatory responses. This schematic serves as a conceptual summary rather than a quantitative model. Created in BioRender; n = 3 independent biological replicates for in vitro assays; n = 5 per group for in vivo bacterial counts. *p < 0.05, p < 0.01, ***p < 0.001, ****p < 0.0001. ns indicates not significant.

Moreover, GNP‐D consistently outperformed its GNP‐L counterpart at the same induction level. Additionally, independent studies report a similar trend: GNP‐D exhibits stronger antibacterial activity than GNP‐L in S. aureus, and even in Escherichia coli [24, 25]. Although these studies differ in nanoparticle geometry and surface chemistry, they report a consistent trend favoring D‐enantiomeric surfaces. To gain mechanistic insight, we next quantified bacterial ATP levels and intracellular ROS. ATP supplies energy for bacterial biosynthesis, nutrient uptake and transport, motility, and stress responses [26]. Its depletion indicates energetic collapse and growth arrest. Luminescence assays showed marked ATP decreases in 1L/1D in both S. aureus and MRSA (Figure 2C). GNP‐D groups were consistently lower than GNP‐L groups. In S. aureus, ATP in 1L was about 2.7‐fold higher than in 1D. We further assessed ROS, a well‐recognized antibacterial mechanism that damages membranes, proteins, and nucleic acids and disrupts redox homeostasis [27]. GNP‐D groups showed higher ROS, with the strongest effect in 1D (Figure 2D). In S. aureus, 1D exceeded 0.1D by 20%. At the 2.5 dose, the GNP‐D group was 17% higher than the L group. These results suggest that D chirality amplifies oxidative stress, especially in 1D, consistent with the observed growth limitation. Building on the ATP and ROS results, we next examined baseline phenotypes linked to biofilm survival in S. aureus and MRSA. Crystal violet assays showed stronger biofilm inhibition in GNP‐D groups, with 1L/1D being the most effective. In S. aureus, 1D inhibited biofilm formation 1.5‐fold more than 1L. In MRSA, 1D exceeded 1L by 1.1‐fold (Figure S4). Thus, chirality modulates biofilm formation, and the pattern mirrors our earlier findings. To assess membrane damage, we measured protein leakage (Figure S5). GNP‐D groups showed higher leakage in both S. aureus and MRSA, consistent with the ROS trends. Overall, these data rule out concentration as a confounder, confirm the superior performance of 1L/1D, and support a genuine chirality‐driven antibacterial effect independent of other physicochemical variables. Despite the inherent complexity of isolating absolute causality, the stark contrast in biological activities between these physicochemical twins points to a clear chirality‐dependent mechanism. This conclusion is systematically validated across multiple scales by our transcriptomic, computational, and genetic knockout data.

2.2.2. In Vitro Modulation of Macrophage Inflammatory Responses by GNP‐L and GNP‐D

In the host infection microenvironment, the efficacy of antibacterial materials depends not only on direct bacterial killing but also on the regulation of innate immune responses [28]. Macrophages act as major effectors of pathogen clearance and key regulators of inflammation [29]. After confirming that GNP‐D exhibits stronger direct antibacterial activity than GNP‐L, we examined whether surface chirality also modulates macrophage inflammatory responses and contributes to overall anti‐infective efficacy. To address this question, we established a RAW264.7 macrophage and bacteria co‐culture model (Figure 2E). RAW264.7 cells were first co‐incubated with S. aureus for 2 h to mimic the early stage of infection. Cells were then treated with either GNP‐L or GNP‐D. Samples were collected at 6 h and 24 h for RT‐qPCR analysis. At 6 h, GNP‐D treatment significantly increased the transcription of Il1b and Il10 compared with GNP‐L, indicating stronger early immune activation (Figure 2F,G). At 24 h, Il1b expression remained high in the GNP‐L group but was markedly reduced in the GNP‐D group to approximately 5% of the GNP‐L level. In contrast, Il10 expression in the GNP‐D group remained significantly elevated and was about 2.5‐fold higher than that in the GNP‐L group. This temporal pattern reflects an early pro‐inflammatory activation followed by a timely shift toward immune regulation and resolution in the GNP‐D group. Together with the observed ATP depletion, increased ROS generation, and inhibition of biofilm formation, these findings demonstrate that GNP‐D achieves superior anti‐infective efficacy by combining effective bacterial killing with balanced immune modulation.

2.2.3. In Vivo Therapeutic Efficacy of GNP‐L and GNP‐D in an Infected Wound Model

The in vitro antibacterial phenotypes were established and showed consistent, reproducible patterns. We next evaluated their efficacy in vivo. Eight‐week‐old female C57BL/6 mice received full‐thickness dorsal wounds inoculated with S. aureus. Based on in vitro results, we applied GNP‐L or GNP‐D (1L/1D) as local rinses on post‐operative days 1 and 3, and collected wound swabs on day 9 for culture (Figure 2H). Bacterial counts showed stronger suppression in the D group (Figure 2I). This agreed with the in vitro trend and indicates activity in a physiological setting. For histological evaluation, Masson's trichrome staining of day‐9 skin showed more continuous and organized collagen fibers in the D‐treated wounds compared with the L‐treated group (Figure 2J). This collagen architecture is indicative of improved tissue remodeling. These histological features are consistent with a transition toward resolution, in line with the in vitro immunomodulatory trends. Together with the in vitro findings, these in vivo results support a chirality‐dependent enhancement of anti‐infective efficacy by GNP‐D. This effect is consistent with the in vitro observations of ATP depletion, elevated ROS generation, membrane perturbation, and biofilm inhibition, which collectively contribute to bacterial growth limitation (Figure 2K).

2.3. Chirality‐Dependent Disruption of Bacterial Structure and Dynamics

2.3.1. Disruption of the Bacterial Cell Envelope and Biofilm Formation

In S. aureus and MRSA, a thick, cross‐linked peptidoglycan reinforced by wall teichoic acids provides a strong mechanical and diffusional barrier [30]. A robust biofilm program further shields cells, promoting persistence and treatment failure. This is precisely the challenge in treating infections caused by these two bacteria. After establishing antibacterial efficacy, we sought morphological evidence of bacterial damage. Specifically, we examined whether the cell envelope was compromised and characterized the modes of breach. This approach enables a more mechanistic understanding of chirality‐dependent injury pathways and their progression. At the microstructural level, SEM compared the strongest conditions (GNP‐1L/1D) with the vehicle control (CON) and showed fewer surviving S. aureus cells at low magnification (Figure 3A). At high magnification, nanoparticles enveloped cells and contact sites showed cell‐envelope distortion and stretching (Figure S6). In 1D, cells showed obvious breaks with contents leaking out, pointing to stronger contact‐based damage under D chirality. In MRSA, we also saw particles wrapping the cells, fewer intact cells, and deformed shapes, but these were less pronounced than those in S. aureus under the same conditions (Figure 3B). At a finer ultrastructural level, we examined cells by TEM. GNP‐treated bacteria showed cell envelope damage of varying severity, including discontinuities, loss of wall material, and cytoplasmic leakage (Figure S7). Arrowheads mark regions of wall loss, and dashed lines indicate interrupted wall continuity (Figure 3C). These morphological changes are consistent with severe disruption of the cell envelope, which would impede growth and viability [31, 32]. The effect is stronger for GNP‐D, with the most pronounced wall damage. Prior studies report that GNPs can compromise Gram‐positive envelopes by perturbing membrane potential, increasing permeability, and causing damage. These are consistent with our observations. We next assessed functional phenotypes. Live/dead staining (DAMO; N‐dimethylaniline N‐oxide, labeling metabolically active cells/PI) showed higher PI‐positive fractions after 1D in both species (Figure 3D,E); quantitatively, the dead/live ratio in S. aureus 1D was 1.8‐fold that of 1L (Figure 3G), and in MRSA 1D was 1.2‐fold that of 1L (Figure 3I). Three‐dimensional confocal imaging demonstrated reduced biofilm biomass and mean thickness after chiral GNPs, with stronger inhibition in 1D (Figure 3F). In S. aureus, the mean fluorescence intensity in 1D decreased by 80.4% versus CON group. Together, these data show that chiral GNPs exert direct, contact‐mediated antibacterial effects with clear chirality dependence. In S. aureus, cell envelope integrity and biofilm formation were markedly disrupted, with GNP‐D showing the strongest effects, supporting a clear chirality dependence of antibacterial activity.

FIGURE 3.

Chiral GNPs affect the morphology, viability, biofilm, and motility of S. aureus and MRSA. (A, B) Scanning electron microscopy (SEM) of S. aureus (A) and MRSA (B) treated with CON, 1L, 1D (representative fields; scale bars 4 µm for wide views and 1 µm for magnified insets; pseudo‐coloring applied for clarity). (C) Transmission electron microscopy (TEM) of S. aureus and MRSA treated with CON, 1L, 1D (scale bar 200 nm); dashed lines outline cells, arrowheads indicate envelope defect sites. (D) Live/dead staining of S. aureus and (E) MRSA treated with CON, 1L, 1D (green, DAMO for live; red, PI for dead). (F) Three‐dimensional confocal reconstructions of biofilm (scale bar 100 µm). (G, I) Quantification of live/dead cell ratio in (D) and (E). (H, J) Quantification of biofilm metrics in (F). (K) Total internal reflection fluorescence (TIRF) microscopy workflow for near‐surface single‐cell dynamics under CON, 1L, 1D (S. aureus; incubation 30 min; details in Methods; created in BioRender). (L) Step‐size (displacement) distributions from TIRF trajectories of S. aureus at ∆t = 0.22 s. (M) mean squared displacement (MSD) versus time‐interval curves of S. aureus; numbers on the curves indicate the fitted α exponents. (N) The trajectory distance frequency distribution histograms of S. aureus. Data are representative of at least three independent preparations (n = 3). *p < 0.05, p < 0.01, ***p < 0.001, ****p < 0.0001. ns indicates not significant.

2.3.2. Restriction of Near‐Surface Single‐Cell Motility

Given the morphology and functional readouts pointing to cell envelope damage, ATP depletion, and elevated ROS, we asked whether these insults manifest in single‐cell motility. We aimed to quantify chirality‐dependent damage via single‐cell dynamics. This approach links ultrastructural injury and energetic stress to biofilm initiation and maturation, using motility metrics as the readout [33, 34]. We used mid‐log S. aureus in Luria‐Bertani broth (LB), because cells are most responsive in this growth phase [10]. Cultures were incubated with GNP‐L or GNP‐D for 30 min, stained with DiD to avoid nanoparticle spectral interference (Figure S8), and imaged by TIRF. Ten‐minute movies were converted to single‐cell trajectories for displacement statistics (Figure 3K). Raw TIRF movies were background subtracted, subpixel localized, drift corrected, and linked into quality filtered trajectories (Figure S9). To probe instantaneous motion, we quantified near‐surface motility by calculating the distribution of frame‐to‐frame displacements at fixed lag times (Δt) from TIRF trajectories (self–van Hove function). The probability of large displacements was lowest in 1D, intermediate in 1L, and highest in control (Figure 3L), indicating that D‐chirality suppresses long‐range near‐surface jumps and yields stronger interfacial confinement. We then fit ensemble mean squared displacement (MSD) to a power law, MSD(Δt) = 4D·Δt^α (Figure 3M). The α exponent was lowest in 1D, 0.337, and higher in 1L, 0.466. The lower α in 1D implied stronger confinement/obstacle interactions and reduced effective mobility. This pattern was consistent with energy depletion and enhanced wall/membrane coupling, though MSD alone did not establish causality. Distance histograms supported this trend: 1D curve was taller and narrower, concentrating probability at short step sizes and suppressing long displacements, whereas 1L curve was lower and broader, indicating a wider spread toward larger steps (Figure 3N). Thus, GNP‐D most strongly restrict near‐surface motion, L‐chirality shows intermediate restriction, and untreated cells remain the most motile. These biophysical readouts align with bulk assays showing ATP decreases, ROS increases, and biofilm inhibition, supporting chirality‐specific modulation of bacterial activity.

2.4. Chirality‐Dependent Transcriptomic Reprogramming of Carbon Transport and Metabolism

Building on the observed chirality signatures in ATP, ROS, motility, and biofilm, we then examined how these outcomes were reflected at the transcriptional level by profiling the S. aureus transcriptome after short‐term exposure (30 min) to GNP‐L or GNP‐D. This analysis revealed perturbed pathways and regulators, including PTS transport, carbohydrate and energy metabolism, envelope maintenance, and stress responses. This transcriptomic analysis linked nanoscale material cues to genome‐wide regulatory programs and provided mechanistic context for the stronger activity of GNP‐D. To probe early effects of chiral GNPs, mid‐log cultures (∼14 h) were exposed to GNP‐L or GNP‐D for 30 min and processed for RNA‐seq (Figure 4A). This window was chosen to capture primary transcriptional responses and to limit secondary metabolic rewiring. Prior to downstream analyses, we first assessed sample quality. Within‐group replicates (L1‐ L3, D1‐D3) showed similar whole‐genome density profiles (Figure 4B), and principal component analysis (PCA) likewise showed high within‐group similarity and stability (Figure S10). These results indicated minimal between‐sample variability and high sequencing quality, supporting the robustness of subsequent analyses. To survey global transcriptional changes, we used a volcano plot and identified many differentially expressed genes (DEGs) (Figure 4C). We detected 61 upregulated genes and 52 downregulated genes. The balanced counts indicated bidirectional transcriptomic reprogramming rather than a uniform on–off response. We then examined enrichment to connect significant differences with phenotypes. Gene Ontology (GO) analysis highlighted amino acid transport, the tricarboxylic acid cycle, and the phosphoenolpyruvate‐dependent sugar phosphotransferase (Figure 4D). The phosphoenolpyruvate‐dependent sugar phosphotransferase term included seven genes, supporting a shift in transport and metabolic routing. Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis of downregulated pathways emphasized carbohydrate and amino acid metabolism. The PTS ranked among the top pathways based on gene count (Figure 4E). Five PTS genes were enriched, indicating pathway‐level enrichment within this module. Together, these enrichments pointed to metabolism and membrane transport as key targets of chiral GNPs, with a particular emphasis on carbon uptake and phosphate handling. Notably, the transcriptional changes were selective rather than globally suppressive, with specific enrichment of transport and carbon metabolism pathways. To link genes with function, we built a GO gene chord diagram and selected representative DEGs based on statistical significance, effect size, and contribution to enriched terms (Figure 4F). We focused on the downregulated transport/metabolic gene mtlA (SAOUHSC_02400; SA_02400). Among PTS‐related genes, mtlA exhibited the largest fold change and the most consistent downregulation across biological replicates, supporting its prioritization as a representative and potentially rate‐limiting node. Reduced uptake and flux were consistent with our observed phenotypes of ATP depletion, constrained motility, and decreased biofilm formation, making such nodes more likely to represent functional bottlenecks rather than secondary stress markers [35]. Within the downregulated set, SA_02400 showed the strongest difference: expression in the L groups was 2.2‐fold higher than in the D groups, indicating clear suppression under D chirality. SA_02400 encodes the mannitol‐specific PTS transporter MtlA, a membrane complex that couples phosphoenolpyruvate‐driven phosphorylation to mannitol import [36]. This directly linked the signal to the enriched categories PEP‐dependent sugar PTS and transmembrane transport. The heatmap (Figure 4G) showed a pronounced and consistent decrease of SA_02400 across GNP‐D replicates. Functionally, SA_02400 downregulation implied reduced mannitol uptake via coupled phosphorylation, limiting carbon entry and the supply of precursors for matrix and biomass.

FIGURE 4.

Transcriptomic responses of S. aureus to chiral GNPs. (A) Schematic illustration of the pre‐RNA‐seq workflow. Mid‐log cultures of S. aureus (LB, 37°C, 200 rpm; 14 h) were exposed for 30 min to GNP‐L, GNP‐D; three biological replicates per condition. Created in BioRender. (B) Global expression distributions for L1‐L3 (GNP‐L) and D1‐D3 (GNP‐D) replicates (log2 normalized expression). n = 3 biological replicates per group (L1‐L3, D1‐D3). (C) Volcano plot of differentially expressed genes (DEGs) comparing GNP‐D and GNP‐L treatments (cutoffs: |log₂FC| ≥ 1, Q value < 0.05). (D) Gene Ontology (GO) enrichment analysis of DEGs, highlighting metabolic and transport‐related processes. (E) Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment of downregulated DEGs. (F) GO chord diagram linking representative DEGs to enriched terms (DEGs selected by |log2FC| and membership in top GO terms). (G) Heatmap of significantly altered genes across L1‐L3, D1‐D3. (H) Schematic summary: chiral GNPs disrupt envelope integrity and metabolic homeostasis, with ROS accumulation, ATP depletion, and transcriptional reprogramming, consistent with impaired biofilm formation and bacterial growth (conceptual illustration). Created in BioRender.

These findings are consistent with a coherent mechanism (Figure 4H): chiral GNPs contact the bacterial envelope and induce cell‐envelope stress (CES), increase ROS, and reduce ATP. The ensuing stress and energy limitation are consistent with selective transcriptional reprogramming rather than global suppression. At the membrane, the mannitol‐specific PTS transporter MtlA is engaged; mannitol import and flux decline, reducing coupled phosphate (PO₄ ^{2 −}) transfer and downstream glycolysis. Consistent with this axis, FPKM (fragments per kilobase of transcript per million mapped reads) comparisons of MtlA/PTS genes showed ptsH highest in both groups, consistent L‐over‐D elevation for mtlA, mtlF, and mtlD, and low SAOUHSC_02974 (Figure S11). The metabolic shortfall limited biomass production and biofilm formation and suppressed growth. Overall, envelope perturbation plus MtlA/PTS throttling linked surface chirality to impaired energy metabolism and diminished biofilm and growth. While the transcriptomic findings remain associative, their strong concordance with TIRF motility, ATP depletion, and ROS levels provides robust phenotypic support for the involvement of the MtlA/PTS axis in chiral‐selective interactions. Together, these data supported the MtlA/PTS axis in carbon metabolism as a central readout of chiral GNP interactions with membranes and proteins, with GNP‐D showing the stronger effect (Figure 4H).

2.5. Chirality‐Dependent Interfacial Coupling of GNPs to the MtlA/PTS Transport Axis

To provide physical evidence linking transcriptomic changes to observed phenotypes, we employed MD to quantify structural and dynamical changes at the protein‐membrane interface [37, 38]. In this way, MD provided physics‐based constraints that complemented bulk assays and RNA‐seq. A protein‐protein interaction network confirmed high‐confidence associations of MtlA with key PTS components (Figure 5A), reinforcing its selection as a representative target. We then constructed a simulated nanoparticle model (Figure S12) and an S. aureus‐mimetic membrane system to calibrate the orientation and environment of MtlA (Figure 5B) [39]. MtlA remained stably embedded in the bilayer, providing a physically reasonable baseline for assessing GNP‐induced effects [40]. Computational models were constructed using L‐/D‐GNP surfaces incorporating asymmetric gold lattice distortions. Thus, simulated interactions reflect the synergistic effects of ligand‐induced surface reconstruction and chiral interfacial organization, rather than intrinsic crystallographic core chirality. To compare how the two chiral GNPs approached MtlA, we examined contact geometry and membrane/protein deformation [41].

FIGURE 5.

Molecular dynamics of chiral nanoparticle and membrane protein systems. (A) Protein–protein interaction network centered on MtlA generated from STRING analysis; node size indicates interaction degree, edge thickness reflects confidence, and colors denote interaction evidence types. (B) Initial membrane‐embedded model of MtlA in a S. aureus–mimetic bilayer composed of POPE, POPG, and TOCL1 (inner and outer leaflets indicated). (C) Representative MD snapshots of GNP‐L and GNP‐D systems at 0, 50, and 100 ns, illustrating nanoparticle approach and local membrane deformation. (D) Time evolution of the GNP‐MtlA center‐of‐mass distance for GNP‐L and GNP‐D systems. (E) Interaction energetics between GNPs and MtlA, including total nanoparticle‐protein potential energy and short‐range Lennard‐Jones interactions (LJ(SR)). (F) Per‐residue root‐mean‐square fluctuation (RMSF) profiles of MtlA. (G) Time‐resolved solvent‐accessible surface area (SASA) of MtlA. (H) Secondary‐structure fractions of MtlA over time derived from DSSP analysis. (I) DSSP residue–time maps for MtlA residues 300–500 in GNP‐L and GNP‐D systems. (J) Structural context of functionally relevant residues Cys425 and His487 in MtlA; insets show local environments with representative heavy‐atom distances (Å). (K) Time‐resolved SASA and inter‐residue distance metrics for the Cys425/His487 region in GNP‐L and GNP‐D systems. (L) MD simulations of FhuB embedded in a lipid bilayer, showing representative configurations with GNPs and corresponding nanoparticle‐protein interaction potentials (upper right) and per‐residue RMSF profiles (lower right). (M) MD simulations of MntB under identical conditions, with representative configurations, nanoparticle‐protein interaction potentials (upper right), and per‐residue RMSF profiles (lower right).

Both systems established near‐surface contact, but GNP‐D induced more pronounced membrane indentation and outward domain displacement at 50 and 100 ns (Figure 5C). Quantitative analyses confirmed this stronger interfacial coupling. MtlA backbone root‐mean‐square deviation (RMSD) plateaued by ∼30 ns in both trajectories, with the GNP‐D system stabilizing at a higher level, reflecting tighter interfacial association rather than global destabilization (Figure S13). This closer proximity was further evidenced by a shorter average GNP‐MtlA center‐of‐mass distance in the D system (Figure 5D). Consistently, GNP‐D exhibited stronger short‐range Lennard‐Jones interactions and a larger decrease in potential energy (Figure 5E), whereas electrostatic terms remained comparable between enantiomers (Figure S14). Driven by this enhanced nanoscale contact, GNP‐D induced spatially confined increases in local protein flexibility (root‐mean‐square fluctuation; RMSF, Figure 5F) and larger temporal fluctuations in solvent‐accessible surface area (SASA) (Figure 5G). Despite these local dynamics, the overall secondary structure remained stable (Figure 5H,I), indicating localized remodeling rather than global unfolding. Specifically, GNP‐D induced distinct localized conformational remodeling at Cys425 and His487, which lie along MtlA's predicted substrate translocation pathway (Figure 5J) [42, 43]. Time‐resolved tracking of these a priori selected sites revealed increased SASA and shortened inter‐residue distances under D‐chirality (Figure 5K), demonstrating micro‐rearrangements that substantially enhanced conformational coupling.

To probe generality versus selectivity, we chose FhuB, the inner membrane channel for siderophore‐bound iron in an ABC importer, and MntB, the membrane subunit of the manganese ABC uptake system; both are central to S. aureus metal acquisition under nutritional immunity (Figure S15) [44]. In stark contrast to MtlA, GNP interactions with both FhuB and MntB were weak, transient, and showed minimal chirality dependence. For both transporters, short‐range interaction energies and electrostatic terms largely overlapped between the L and D trajectories (Figures S17 and S20). Furthermore, FhuB and MntB rapidly reached stable equilibration without global remodeling (RMSD, Figures S16 and S19), exhibiting nearly superimposable flexibility profiles (RMSF, Figure 5L,M) and maintaining their transmembrane helical cores (DSSP, Figures S18 and S21). Overall, the lack of sustained conformational remodeling in these control proteins confirms that the strong, stereospecific interfacial coupling driven by surface chirality is selectively targeted toward the MtlA/PTS axis.

2.6. MtlA‐Dependent In Vivo Efficacy of Chiral GNPs in an Infected Wound Model

To biologically validate the MD predictions, we evaluated an MtlA‐deficient strain (ΔmtlA; RH). Successful deletion was confirmed by PCR and Sanger sequencing (Figure S22), and functionally validated on mannitol salt agar, where RH formed pink colonies—indicating impaired mannitol utilization—unlike the yellow wild‐type (S. aureus; SA) colonies (Figure S23). A murine dorsal wound infection model was subsequently established (Figure 6A). Serial photographs (Figure 6B) and quantitative analysis (Figure 6C) demonstrated that SA+D significantly accelerated wound contraction relative to SA and SA+L, approaching the uninfected control (CON) by day 9. Crucially, in the RH background, both GNPs improved healing equivalently without chiral separation (RH+L ≈ RH+D), confirming that the stereospecific advantage relies on the MtlA/PTS axis. Furthermore, systemic biosafety was confirmed by stable body weights (Figure 6D) and normal major organ histology (Figure S24).

FIGURE 6.

MtlA‐dependent in vivo efficacy of chiral GNPs in a murine wound infection model. (A) Schematic illustration of the full‐thickness dorsal wound infection model and treatment timeline. Wounds were infected with SA or RH on day 0, followed by local treatment with PBS (CON/SA/RH), GNP‐L (SA+L/RH+L), or GNP‐D (SA+D/RH+D) on days 1 and 3. Animals were monitored and harvested on day 9. Created with BioRender. (B) Representative photographs of wound sites from different groups at Days 0, 1, 4, 7, and 9 post‐infection, with corresponding wound‐area segmentation shown below. Scale bar: 3 mm. (C) Quantification of remaining wound area ratio over time based on standardized photographs. (D) Body weight changes of mice in each group during the experimental period, indicating systemic tolerability. (E) Representative agar plates showing residual bacterial burden in wound swabs collected on day 9. (F) Quantification of relative bacterial survival based on CFU counts in (E). (G) Representative H&E‐stained sections of wound tissues collected at day 9. Yellow asterisks indicate the areas of inflammatory cell infiltration Scale bars, 100 µm; n = 5 mice per group for wound closure analysis; n = 3 for histological quantification.

To quantify residual bacterial burden, day 9 swab cultures showed SA+D with minimal residual bacterial burden, approaching the uninfected control (CON) (Figure 6E,F), consistent with effective infection control; SA+D showed 31.1% lower survival than SA+L. By contrast, in the RH background the overall burden was lower than SA, consistent with impaired mannitol utilization. However, the chiral contrast largely collapsed: RH+D was only 16.9% lower than RH+L, indicating weak chiral responsiveness and supporting an MtlA‐linked component. Skin H&E paralleled these findings. SA+D displayed a more complete epithelium with less inflammation and necrosis. SA+L showed moderate improvement. SA and RH retained inflammatory damage (Figure 6G). Masson's trichrome staining (Figure S25) revealed more continuous and densely organized collagen bundles in SA+D, whereas SA+L showed intermediate remodeling and SA and RH displayed fragmented collagen with interstitial edema. This pattern indicates more advanced tissue remodeling in SA+D, consistent with reduced inflammation and accelerated wound maturation.

2.7. MtlA‐Dependent Immune Remodeling and Motility Suppression by Chiral GNPs In Vivo

2.7.1. MtlA Mediates GNP‐D‐Driven Reprogramming of Wound Immune and Inflammatory Pathways

In vivo, GNP treatment accelerated wound closure and reduced histologic inflammation, most prominently with GNP‐D. To determine if this involved active immune remodeling, we evaluated day‐9 wound skin. Immunohistochemistry revealed the lowest CD86 and highest CD206 expression in SA+D (Figure 7A), indicating attenuated inflammatory activation and enhanced repair signatures. Recognizing the broad stromal distribution of these markers in murine skin, we interpreted them as functional activation readouts and complemented this with flow cytometry [45].

FIGURE 7.

Immune profiling, metabolic readouts, and near‐surface dynamics. (A) Representative IHC staining of CD86 and CD206 in day‐9 wound sections. (B) Flow‐cytometry schedule (infection day 0; treatments days 1 and 3; harvest day 5). (C) UMAP (uniform manifold approximation and projection) embeddings of wound single‐cell suspensions colored by IFN‐γ or I‐A/I‐E signal. (D, E) Radar summaries of adaptive/innate subsets and cytokine‐positive fractions (gating in Methods). (F) Schematic summary of immune remodeling and biofilm attenuation. (G) Intracellular ATP and (H) ROS levels in SA, RH, RH+L, and RH+D groups. (I) Representative near‐surface tracking snapshots (50 ms). (J) Displacement distributions of SA and RH. (K) MSD of SA and RH with fitted α indicated. (L) The trajectory distance frequency distribution histograms of SA and RH; n = 5 mice per group for flow cytometric profiling. *p < 0.05, p < 0.01, ***p < 0.001, ****p < 0.0001. ns indicates not significant. The flow‐cytometry gating strategy is shown in Figure S27.

To capture early dynamics, we profiled day‐5 wounds (Figure 7B) with all files passing FlowAI QC (quality control) (Figure S26). Unsupervised UMAP (uniform manifold approximation and projection) and t‐SNE (t‐distributed stochastic neighbor embedding) embeddings revealed that SA+D distinctly reduced IFN‐γ–associated signals while expanding the MHC‐II (I‐A/I‐E)^high APC compartment (Figure 7C, Figure S28). Quantitative summaries confirmed early reorganization in the SA background (Figure 7D,E): SA+D strongly suppressed IFN‐γ–producing T cells, reduced neutrophil/B‐cell infiltration, and expanded APCs, signaling a shift from acute inflammation toward resolution. Crucially, this coordinated immunomodulation was largely abolished in the MtlA‐knockout (RH) background, where RH+D and RH+L exhibited negligible differences (Figure 7F). Together, these data demonstrate an MtlA‐dependent, two‐phase immune trajectory in wild‐type infection: an early clearance phase (day 5) with restrained inflammatory output, followed by a late remodeling phase (day 9) facilitating tissue repair.

2.7.2. MtlA Knockout Abolishes Chiral Separation in Metabolic and Near‐Surface Kinetic Responses

To definitively link the chiral advantage to MtlA, we evaluated metabolic and kinetic readouts in the RH strain. While GNPs still induced non‐specific metabolic stress, chiral separation was lost: RH+L and RH+D exhibited nearly identical ATP depletion (3.3% difference, Figure 7G) and comparable ROS elevations (Figure 7H).

This convergence extended to single‐cell near‐surface motility (Figure 7I). TIRF tracking revealed nearly overlapping displacement distributions (Figure 7J), equivalent confinement with identical anomalous diffusion exponents (Δα = 0.03, Figure 7K), and similar step‐length distributions between RH+L and RH+D (Figure 7L). Although baseline nanoparticle toxicity persisted, the striking loss of L–D separation across all metabolic, single‐cell dynamic, and immunological readouts in the knockout strain provides robust phenotypic evidence that the in vivo stereospecific advantage of GNP‐D is predominantly mediated by the MtlA/PTS axis. Although the ΔmtlA results provide strong genetic evidence for MtlA involvement, complementation of MtlA in the mutant background remains necessary to further confirm causality and exclude secondary effects associated with altered bacterial metabolism or membrane physiology.Future studies incorporating MtlA complementation, site‐directed mutagenesis, protein‐ligand binding affinity assays, carbon source modulation, and combination therapy will further elucidate the molecular nuances of this chiral‐selective metabolic interference [2, 46].

3. Conclusions

Overall, this work establishes surface chirality as an independent and functionally relevant design dimension for antibacterial nanomaterials. By rigorously controlling physicochemical parameters, we demonstrate that GNP‐D consistently outperforms its L‐enantiomer in suppressing bacterial viability, biofilm formation, and near‐surface motility. Mechanistically, we identify the mannitol‐specific transporter MtlA as a stereospecific molecular target. Molecular dynamics simulations reveal that GNP‐D facilitates tighter interfacial coupling and localized structural perturbations at the MtlA active site, while genetic ablation of mtlA (RH strain) markedly attenuates the D/L performance gap across metabolic, kinetic, and immune readouts. These findings provide robust genetic and phenotypic support for an MtlA‐mediated chiral‐selective mechanism. In vivo, GNP‐D effectively clears infections and promotes a coordinated shift toward a pro‐repair immune microenvironment with a favorable safety profile. This study introduces a “chirality–membrane protein coupling” strategy, positioning chirality as a generalizable tool for the targeted modulation of bacterial metabolism and host‐pathogen interactions across diverse biomedical platforms.

4. Methods

4.1. Bacterial Strains and Culture Conditions

S. aureus (ATCC 35556) and methicillin‐resistant S. aureus (MRSA; ATCC BAA‐1717) were grown overnight in LB at 37°C with shaking. Cultures were diluted 1:100 into fresh LB and grown to mid‐log phase (OD₆₀₀ = 0.6–0.8) before use, unless otherwise stated.

4.2. Mammalian Cell Culture and Viability Assay

NIH/3T3 (ATCC CRL‐1658), L929 (ATCC CCL‐1), and RAW264.7 murine macrophages were maintained under standard culture conditions at 37°C in a humidified atmosphere containing 5% CO₂. NIH/3T3 and L929 cells were detached using trypsin at 37°C for 3 min and collected by centrifugation (1000 rpm, 5 min). Cells were seeded into 96‐well plates (100 µL per well; 1 × 10⁵ cells mL^{− 1}) and allowed to adhere overnight. For cytocompatibility assessment, cells were exposed to gold nanoparticles (GNPs; 100–500 µg mL^{− 1}; vehicle control included) for 4 h at 37°C. After treatment, supernatants were removed, and cells were washed three times with PBS. Fresh culture medium containing 10% CCK‐8 reagent (100 µL per well) was added and incubated for 2 h at 37°C. Reaction mixtures were transferred to fresh plates, and absorbance at 450 nm was measured using a microplate reader. Cell viability was expressed relative to vehicle‐treated controls.

4.3. Intracellular ROS Measurement in Bacteria

Intracellular reactive oxygen species (ROS) in SA, MRSA, and the RH were measured using a DCFH‐DA ROS Assay Kit (Beyotime, S0033S) according to the manufacturer's protocol. Mid‐log cultures were adjusted to OD₆₀₀ = 0.6–0.8, harvested, and resuspended to ∼1 × 10⁶–2 × 10⁷ cells mL^{− 1}. Cells were loaded with DCFH‐DA (10 µM) at 37°C for 20–30 min in the dark with intermittent mixing, followed by three washes with PBS to remove excess probe. Probe‐loaded cells were treated with GNP‐L or GNP‐D (typically 100 µg mL^{− 1}) for the indicated times. Rosup was added only to designated positive‐control wells. Fluorescence of oxidized DCF was recorded using a microplate reader (Ex 488 nm/Em 525 nm). Background signals from probe‐only and “GNP + probe without bacteria” controls were subtracted.

4.4. Bacterial ATP Quantification

Mid‐log cultures were normalized by OD₆₀₀ and lysed on ice using ATP lysis buffer (Beyotime, S0026) according to the manufacturer's protocol. Lysates were clarified by centrifugation (12 000 g, 5 min, 4°C). ATP standards were prepared in parallel. ATP working solution (100 µL) was dispensed into white 96‐well plates and equilibrated at room temperature for 3–5 min. Samples (20 µL) were added, mixed rapidly, and luminescence was recorded immediately. Values were blank‐corrected and quantified using a standard curve. Unless otherwise specified, ATP values were normalized to total protein content (BCA) and/or to OD₆₀₀ as stated in figure legends.

4.5. Bacterial Protein Leakage

Bacteria were set to OD₆₀₀ ≈ 0.6, washed, resuspended in PBS, and incubated with GNPs according to the experimental schedule. After treatment, cells were removed (4°C, 12 000 g, 5 min) and supernatants were filtered through 0.22 µm membranes. Extracellular protein was quantified by BCA assay (Thermo) with BSA standards; absorbance at 562 nm was measured on a plate reader. Reagent blanks and “GNP without bacteria” controls were included.

4.6. RAW264.7 Infection, L/D‐GNP Treatment, and RT‐qPCR Analysis

RAW264.7 murine macrophages were cultured in Dulbecco's modified Eagle medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and 1% penicillin–streptomycin at 37°C in a humidified atmosphere containing 5% CO₂. Cells were seeded into multi‐well plates and allowed to adhere overnight prior to bacterial challenge. For infection, RAW264.7 cells were cocultured with bacteria at a multiplicity of infection (MOI) of 10 for 2 h. After coculture, the supernatant was carefully removed, and the cells were gently washed to eliminate nonadherent bacteria. Fresh culture medium containing gentamicin was then added to eliminate extracellular bacteria, following a standard gentamicin protection protocol. After gentamicin treatment, the medium was replaced with antibiotic‐free DMEM. Subsequently, cells were treated with GNP‐L or GNP‐D at the indicated concentrations, while control groups received an equivalent volume of vehicle. Cells were further incubated for the designated time period prior to RNA extraction. Total RNA was extracted from cells using TRIzol reagent (Takara, cat. no. 9109) and reverse‐transcribed into cDNA using the RT Master Mix (ABclonal, cat. no. RK20433). RT—qPCR was conducted using SYBR qPCR Master Mix (Vazyme, cat. no. Q312 – 02) on a LightCycler 480 II qPCR system (Roche). All reactions were run in triplicate. Relative gene expression was normalized to the endogenous control GAPDH and calculated using the comparative 2^(‐ΔΔCt) method. The sequences of specific primers are listed in Table S2.

4.7. Crystal‐Violet Biofilm Assay

Sterile glass coverslips were placed in 96‐well plates. An inoculum (200 µL; ∼1 × 10⁵ CFU mL^{− 1}) of SA/MRSA was added per well and cultured for 24 h at 37°C to allow formation of mature biofilms. After gentle PBS washes, biofilms were treated with GNP‐L or GNP‐D (typically 100 µg mL⁻¹) for 24 h. Biofilms were fixed with 2.5% glutaraldehyde (2 h), stained with 0.1% crystal violet (20 min), washed with PBS (×3), air‐dried, and eluted with 30% glacial acetic acid (30 min). Absorbance was read at 570 nm. Antibiofilm rate (%) was calculated as: antibiofilm rate (%) = (OD_control − OD_sample)/OD_control×100%.

4.8. Synthesis and Physicochemical Characterization of Chiral Gold Nanoparticles

4.8.1. Synthesis of Gold Nanoparticles

Chiral GNPs (GNP‐L/D) were prepared by a CTAB‐mediated seed‐growth route adapted from published protocols [17]. Briefly, ∼50 nm Au nanocube seeds were generated by NaBH₄ reduction of HAuCl₄ at 30°C, washed, and redispersed in 1 mM CTAB. Growth solutions (CTAB/HAuCl₄/ascorbic acid; Sigma) containing L‐ or D‐cysteine (0.1 mM; 0.1–5 µL; TCI) were inoculated with 50 µL seeds and incubated at 30°C until a pink‐to‐blue color shift. Products were purified by two low‐speed spins (1700×g, 5 min) and stored in 1 mM CTAB.

4.8.2. SEM, Zeta Potential, UV–vis, and CD Measurements

Dilute suspensions (5–10 µg mL^{− 1}) were drop‐cast on cleaned silicon wafers, air‐dried, rinsed with ultrapure water, and dried in a desiccator. Samples were sputter‐coated with 2–5 nm Pt/Pd and imaged on a field‐emission SEM. Controls included probe‐only blanks, vehicle, and “GNP plus probe without bacteria” to account for autofluorescence/quenching. Unless otherwise noted, GNP‐L/D = 100 µg mL^{− 1}. GNPs were dispersed in low‐ionic‐strength buffer (10 mM NaCl, pH 7.4) to 20–50 µg mL^{− 1}, passed through a 0.22 µm filter, and sonicated for 5 min. Electrophoretic mobility was measured at 25°C in disposable folded‐capillary cells and converted to ζ‐potential. Data are mean ± SD from ≥3 independent preparations. Spectra were collected in 1 cm quartz cuvettes from 200–800 nm (1 nm step; 1 nm bandwidth). Suspensions (10–30 µg mL^{− 1}) were gently vortexed and briefly sonicated immediately before measurement. Each sample was measured in triplicate and reported as mean ± SD. Unless stated, 100 µg mL^{− 1} GNP‐L/D were used for routine characterization with appropriate blanks and controls to correct for background signals. Representative 1L and 1D batches were prepared by adding 1 µL of 0.1 mM L‐ or D‐cysteine to purified GNPs, followed by light centrifugation (7000 g, 60 s) and resuspension in 1 mM CTAB. Suspensions were diluted to reach A ≈ 0.5 near the LSPR and scanned in 1 mm quartz cuvettes at 25°C ± 0.2°C from 400–850 nm (1 nm step, DIT = 1 s, three accumulations). Baselines with 1 mM CTAB were subtracted. The cuvette was rotated by 90° to exclude linear dichroism artifacts. Only regions with A ≥ 0.03 were evaluated.

4.9. Bacterial Morphology Analysis by SEM and TEM

SA and MRSA were grown to mid‐log phase and treated with GNP‐L or GNP‐D (typically 100 µg mL⁻¹) for 2 h. Cells were collected by centrifugation, washed with PBS (×3), and fixed (4% paraformaldehyde for 10 min at room temperature). For SEM, fixed bacteria were dehydrated through graded ethanol, dried, mounted, sputter‐coated, and imaged. For TEM, fixed samples were processed by standard embedding/sectioning procedures and imaged on a transmission electron microscope. Controls included vehicle‐only bacteria and GNP‐only blanks where applicable.

4.10. RNA Sequencing and Analysis

SA cultures were treated with GNP‐D or GNP‐L and processed as three biological replicates per condition; pellets were collected by centrifugation and stored at −80°C for subsequent analysis. Library preparation and sequencing were performed by BGI using standard protocols for the DNBSEQ platform. In total, six samples were sequenced, yielding on average 4 Gb data per sample. Reads were aligned to the reference genome using the provider's standard pipeline. Gene annotation was performed against the NCBI Genome database. Data visualization was performed in R [47], and heatmap was plotted by https://www.bioinformatics.com.cn, an online platform for data analysis and visualization [48].

4.11. PCR, Agarose Gel Electrophoresis, and Gene Knockout Construction

Transformed bacteria were cultured to mid‐log phase. Bacterial RNA or plasmid DNA was used as template; for RNA, cDNA was synthesized in 20 µL reactions following the reverse transcriptase manual (Abclonal, Cat. RK20429; 25°C 10 min, 50°C 15–30 min, 85°C 5 min, hold at 4°C). PCR products were separated by horizontal agarose gel electrophoresis on 1% gels prepared with TAE buffer (50× stock; Biosharp), alongside a 250‐bp DNA marker (Tsingke). Bands were imaged under blue light, and fragment sizes were assigned by comparing migration distances to the ladder (Primers are listed in Table S2). A temperature‐sensitive Staphylococcus shuttle plasmid pKOR1 (#133446, Fenghbio) was used to generate an in‐frame deletion of mtlA (SAOUHSC_02400) in the RH. Genomic DNA from RH served as PCR template to amplify 1.0 kb upstream (UP) and downstream (DN) homology arms flanking the mtlA coding sequence. The UP and DN fragments were fused by overlap‐extension PCR to create an allelic‐exchange cassette lacking the entire mtlA ORF, leaving a small scar without polar effects (Primers are listed in Table S2). The fused fragment was cloned into pKOR1 (#133446, Fenghbio), transformed into a ccdB‐tolerant E. coli cloning strain, and verified by restriction mapping and Sanger sequencing with vector‐ and locus‐specific primers. The verified plasmid was moved into S. aureus RN4220 (restriction‐deficient intermediate) by electroporation at the permissive temperature (30°C) with chloramphenicol selection (Cm). Plasmid DNA recovered from RN4220 was then electroporated into the RH parent (0.2 cm cuvette; ∼2.3 kV, 25 µF, 100 Ω; cells washed in 0.5 M sucrose), recovered for 1–2 h at 30°C in TSB + 0.5 M sucrose, and plated on Cm to obtain primary transformants carrying pKOR1‐mtlA. For chromosomal integration (single crossover), Cm‐resistant transformants were passaged at the non‐permissive temperature (42°C–43°C) on Cm plates until colonies stably maintaining the integrated plasmid were obtained. Correct integration at the mtlA locus was confirmed by colony PCR using primers annealing outside the UP/DN arms. To force resolution to double crossover, integrants were grown without antibiotic at 30°C and counter‐selected on tryptic soy agar containing 1 M sucrose (sacB counterselection); where applicable, induction of the plasmid's counter‐selectable cassette was used per pKOR1 instructions. Sucrose‐resistant, Cm‐sensitive colonies were screened by diagnostic PCR with external primers flanking the mtlA locus to distinguish wild‐type revertants from ΔmtlA replacements (expected size shift equal to the deleted ORF). Representative positives were screened by diagnostic PCR (Figure S22). Amplicons were analyzed by horizontal agarose gel electrophoresis to confirm the expected size, and the recombination junctions were subsequently verified by Sanger sequencing. The confirmed ΔmtlA mutant was designated RH and cured of any residual plasmid by growth without selection. Loss of mtlA was further validated functionally on mannitol‐salt agar and by qPCR of mtlA transcripts (Figure S23).

4.12. TIRF Imaging and Single‐Cell Tracking

Mid‐log S. aureus and RH were grown in LB at 37°C (OD₆₀₀ ≈ 0.6–0.8), pelleted, and resuspended in PBS. Cultures were incubated with GNP‐L or GNP‐D for 30 min, washed, and membrane‐stained with DiD (far‐red; chosen to avoid spectral overlap with nanoparticle scattering; Figure S8); excess dye was removed and cells were transferred to PBS imaging buffer. Cells were loaded into glass and imaged on a Nikon TIRF microscope with a 100×/1.45 NA oil objective, 640‐nm excitation, and an EMCCD camera. The incident angle was set to give an evanescent‐field penetration depth of ∼100 nm. Time‐lapse movies (∼40×40 µm field) were recorded at 50 ms per frame for 10 min at 25°C–30°C. Movies were background‐subtracted, cells were localized with subpixel precision, and trajectories were linked with a maximum frame‐to‐frame displacement of ∼1.0 µm; stage drift was corrected by cross‐correlation.

4.13. Molecular Dynamics Simulations

MD simulations were used to characterize how chiral gold nanoparticles (GNPs) engage bacterial membrane proteins MtlA, FhuB, and MntB. Protein structures for MtlA (AF‐Q2FW99‐F1‐v4), FhuB (AF‐Q2G1Z2‐F1‐v4), and MntB (AF‐Q2G2D9‐F1‐v4) were obtained from the AlphaFold Protein Structure Database [49]. Enantiomeric GNP models (L‐ vs D‐cysteine‐capped) were built in Materials Studio 2017 (BIOVIA). Protein‐protein interaction (PPI) information was retrieved from STRING (https://string‐db.org) for S. aureus and used to build the PPI network. Protein structures were oriented in a bacterial‐like lipid bilayer (POPE/POPG/TOCL1 mixture) constructed with CHARMM‐GUI, solvated with TIP3 water and neutralized with 150 mM NaCl [50]. All‐atom simulations were performed in GROMACS 2025.2 [51] using CHARMM36m protein/lipid parameters [52, 53] and PME electrostatics (real‐space cutoff 1.2 nm; LJ cutoff 1.2 nm; Verlet neighbor list). Systems underwent steepest‐descent minimization, NVT pre‐equilibration, and NPT equilibration, followed by 100 ns production runs. Trajectories were saved every 10 ps. Conformational stability was assessed by backbone RMSD and per‐residue RMSF of each protein. Secondary‐structure persistence was quantified with DSSP (define secondary structure of proteins) over time; global and per‐residue solvent‐accessible surface area (SASA) were computed (probe radius 0.14 nm), including functional residues Cys425 and His487 of MtlA. To resolve GNP‐protein coupling, non‐bonded interaction energies were decomposed via energy groups and rerun analysis to obtain four components between the GNP and the protein: Lennard‐Jones (short‐range), Coulomb (short‐range), Coulomb (reciprocal/PME), and total potential energy. Protein‐GNP contacts, minimum distances, and time‐resolved SASA of functional sites were extracted from the same trajectories. Unless stated otherwise, analyses used the full 0–100 ns trajectories. Figures report Mean ± SD across independent systems (L vs D for each target), with time‐windowing where indicated to exclude initial equilibration. Unless otherwise specified, GNP‐L/D were used at 100 µg mL^{− 1}. Controls included probe‐only blanks, vehicle, and GNP‐plus‐probe without bacteria to correct autofluorescence/quenching.

4.14. In Vivo SA‐Infected Cutaneous Wound Healing Evaluation and In Vivo Toxicity

The animal experiment protocol was conducted in accordance with national guidelines for the care and use of experimental animals. Female C57BL/6 mice (6 weeks old) were used and randomly divided into seven groups (medium (liquid LB, hopebio) only, S. aureus bacteria, SA+D treatment, SA+L treatment, RH bacteria, RH+D treatment, RH+L treatment). For anesthesia, sodium pentobarbital (20 mg mL^{− 1}, 40 mg kg^{− 1}) was administered intraperitoneally. The dorsal hair was removed using depilatory cream and a razor, followed by disinfection with ethanol to expose the dorsal skin extending from the axilla to the hind limbs. An 8 mm‐diameter circular defect was excised from the dorsal skin. Hemorrhage was controlled by cotton swabs, and wound size was documented by digital imaging with a steel ruler as reference. To establish infection, the wound fascia was evenly punctured with a syringe needle, and 20 µL of S. aureus or RH suspension (4×10⁷ CFU mL⁻¹) was applied to each wound. On days 1 and 3 post‐infection, wounds were treated with 20 µL of GNP‐D or GNP‐L, respectively. Wounds were covered with sterile dressings and secured with bandages to maintain infection stability. Wounds were documented on days 0, 1, 4, 7, and 9, and animals were euthanized on day 9 to collect tissue for subsequent slice analysis including hematoxylin and eosin (H&E, EpiZyme), Masson (Solarbio). Two independent cohorts were used: a flow‐cytometry cohort (terminal day 5) and a histology cohort (terminal day 9), randomized separately. At the study endpoint, mice were euthanized and major organs were collected, fixed in 4% paraformaldehyde, processed to paraffin, and examined by routine H&E to assess systemic toxicity.

4.15. Flow Cytometry

Female C57BL/6 mice (6 weeks) were randomized into seven groups: medium only, SA, SA + GNP‐D, SA + GNP‐L, RH, RH + GNP‐D, RH + GNP‐L. Anesthesia was induced with sodium pentobarbital (20 mg mL^{− 1}, 40 mg kg^{− 1}, i.p.). After dorsal depilation and ethanol disinfection, a full‐thickness 8 mm circular wound was created. Hemorrhage was controlled with sterile swabs, and baseline images were taken with a steel ruler reference. To establish infection, the wound fascia was evenly punctured with a syringe needle and 20 µL bacterial suspension (4 × 10⁷ CFU mL^{− 1} of S. aureus or RH, as assigned) was applied. On days 1 and 3 post‐infection, 20 µL GNP‐D or GNP‐L (vehicle for infection‐only controls) was topically administered. Wounds were covered with sterile dressings and bandaged to maintain a stable microenvironment. On day 5, the wound bed together with a ∼5 mm peripheral rim was excised and immediately immersed in ice‐cold DMEM (Vazyme). Tissue was finely minced and digested at 37°C with gentle shaking (180 rpm) for 60–90 min in a dissociation buffer containing collagenase IV (1 mg mL^{− 1}, Vazyme), DNase I (50 µg mL^{− 1}, Vazyme), and 10% fetal bovine serum. The digest was passed through a 70 µm cell strainer to yield single‐cell suspensions, collected by centrifugation (300–500 g, 10 min, 4°C), washed with PBS (Vazyme), and resuspended in RPMI‐1640 supplemented with 10% FBS. For cytokine accumulation, cells were stimulated with PMA and brefeldin A(Biolegend) at 37°C for 4–6 h. Surface antigens were labeled with fluorophore‐conjugated antibodies in PBS containing 2% FBS for 20 min at room temperature in the dark. After washing, cells were fixed in 4% paraformaldehyde for 20 min and permeabilized; intracellular targets were then stained with the indicated antibodies for 20 min. Samples were washed, resuspended in PBS, and analyzed on a benchtop flow cytometer (Beckman). Antibody panels are listed in Table S1.

4.16. Histology and Immunohistochemistry

Paraffin sections (5 µm) were prepared after 4% paraformaldehyde fixation. For Masson's trichrome, slides underwent deparaffinization/graded rehydration and the standard sequence of Weigert's hematoxylin, Biebrich scarlet–acid fuchsin, phosphomolybdic–phosphotungstic differentiation, and aniline blue; collagen, muscle, and nuclei appeared blue, red, and blue‐black, respectively. For H&E, sections were deparaffinized, rehydrated, stained with hematoxylin (≈5 min), blued, counterstained with eosin Y, dehydrated, cleared in xylene, and resin‐mounted. For IHC, endogenous peroxidase was blocked (3% H₂O₂), antigens were retrieved in citrate buffer (10 mM, pH 6.0; 95°C–100°C, 15–20 min), and slides were blocked with 3% BSA before incubation with primary antibodies (overnight, 4°C) and HRP‐conjugated secondary antibodies; DAB was used for chromogenesis. All slides were counterstained with hematoxylin, dehydrated, cleared, mounted, and imaged under bright‐field microscopy. Isotype and primary‐omission controls were included; antibodies are listed in Table S1.

4.17. Statistics

All experiments were performed with at least three independent biological replicates (n≥3). Data are presented as mean ± SD unless otherwise indicated. Statistical analyses were performed using GraphPad Prism 8.0. Normality and homoscedasticity were verified by Shapiro‐Wilk and Brown‐Forsythe tests, respectively, prior to parametric analysis. Comparisons between multiple groups were conducted using one‐way ANOVA followed by Tukey's post hoc test for multiple comparisons. For time‐course wound‐healing data (Figure 6C), two‐way repeated measures ANOVA followed by Bonferroni's post hoc test was applied to evaluate the effects of treatment and time. Pairwise comparisons between two groups were performed using Student's t‐test where appropriate. A two‐tailed p < 0.05 was considered statistically significant. *p < 0.05, p < 0.01, ***p < 0.001, ****p < 0.0001. ns indicates not significant.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File: smll74522‐sup‐0001‐SuppMat.docx.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.