Model

Introduction & Research Motivation

In the engineering cycle of synthetic biology, modeling serves not merely as a supplement to experiments but as a crucial component that guides design and interprets results. In the context of antimicrobial peptide (AMP) engineering, identifying which amino acid substitutions could positively influence activity is often challenging. Although experimental assays can reveal whether certain mutants exhibit enhanced or diminished activity, they seldom explain why such changes occur or how the underlying mechanisms function. Furthermore, due to biosafety regulations, as a high school team we were unable to directly evaluate our peptides against typical zoonotic pathogens. To overcome these limitations, we employed protein structure prediction and molecular docking to explore the three-dimensional conformations of candidate AMPs and analyze their interactions with bacterial targets in depth.

Our modeling work was conducted around two core scientific questions:

1. Selection and rationality analysis of mutation sites

Building upon the original Ulink-AMP sequence, we predicted which key amino acid substitutions might alter its binding mode with E. coli peptidoglycan glycosyltransferase, thereby improving its antibacterial efficacy. The molecular docking results provided structural evidence explaining why AMP-TB2 exhibited stronger inhibitory activity in experiments.

2. Structural basis of broad-spectrum antibacterial potential

Experimental data showed that AMP-TB2 could inhibit both Gram-positive and Gram-negative bacteria. To verify its broad-spectrum potential, we selected three representative zoonotic pathogens—Staphylococcus aureus, Salmonella, and Campylobacter—and performed molecular docking on their key target proteins. The predicted binding patterns and stability offered theoretical support for the experimentally observed inhibition across multiple bacterial species.

Through this series of efforts, we established a closed-loop framework in which modeling guided experimental design and experimental results in turn validated modeling predictions. The model not only helped us identify the most promising mutants at an early stage—reducing unnecessary trials—but also provided mechanistic explanations for the experimental outcomes. Moreover, the prediction of broad-spectrum potential further expanded the applicability of our model, enabling it to support both mutation optimization and the elucidation of AMP-TB2’s cross-pathogen mechanism of action. This tightly integrated workflow exemplifies the model–experiment integration emphasized in the iGEM Gold Medal criteria.

Data & Pipeline

To ensure the reproducibility of our modeling results, we adhered to the principles of transparency, standardization, and shareability throughout data selection and toolchain development. All input sequences were obtained from publicly accessible databases. The amino acid sequences of the antimicrobial peptide Ulink-AMP and its optimized mutant AMP-TB2 were sourced from the AMPsphere database and our team’s design results. The receptor protein structures were retrieved from the Protein Data Bank (PDB), including E. coli peptidoglycan glycosyltransferase RodA–PBP2 (PDB ID: 8TJ3), Staphylococcus aureus penicillin-binding protein PBP (PDB ID: 3DWK), Campylobacter N-acetyltransferase PseH (PDB ID: 4XPK), and Salmonella glycosyltransferase (PDB ID: 5N80).

These targets all play essential roles in cell wall synthesis or glycosyl modification, making them ideal candidates for predicting the potential mechanisms of AMP–protein interactions.

For toolchain development, we implemented a multi-layered analytical workflow. First, we employed AlphaFold3 to predict the three-dimensional structure of Ulink-AMP and selected high-confidence conformations based on pLDDT and PAE scores. Next, LightDock was used to perform flexible peptide–protein docking, allowing us to better simulate the binding environment of membrane-associated complexes. On this basis, PDBePISA was applied to calculate the interface area and binding free-energy change (ΔG), providing quantitative metrics for assessing the stability and potential functional relevance of the predicted interactions.

In the broad-spectrum analysis, we focused on AMP-TB2, the mutant exhibiting the strongest experimental performance, and further evaluated its binding characteristics across multiple zoonotic pathogens. By comparing parameters such as binding energy, the number of hydrogen bonds and hydrophobic interactions, and interface area, we verified at the structural level whether AMP-TB2 maintains a consistent mode of action across different pathogens. This analysis not only provides a theoretical basis for the experimentally observed broad-spectrum antibacterial activity but also extends the applicability of our model beyond experimental design, demonstrating its value in mechanistic interpretation and predictive evaluation.

Part1-Selection and Rationality Analysis of Mutation Sites

Methods

Ulink-AMP Structure Prediction

We used AlphaFold3 (AF3) to predict the three-dimensional structure of Ulink-AMP. AF3, developed by Google DeepMind and Isomorphic Labs, represents the latest generation of structure prediction models. Building upon AlphaFold2, it integrates evolutionary information and diffusion-based network optimization, enabling simultaneous modeling of proteins, nucleic acids, and small molecules, as well as their interactions. Using this model, we generated multiple candidate conformations and selected high-confidence structures based on pLDDT and PAE scores, which were subsequently used as input for molecular docking analysis.

Molecular Docking Based on LightDock

Literature reports indicate that peptidoglycan (PG) is a key structural component of bacterial cell walls, whose biosynthesis is catalyzed cooperatively by glycosyltransferases (GTs) and transpeptidases (TPs). In E. coli, the RodA–PBP2 complex, formed by RodA (GT) and PBP2 (TP), constitutes a core component of the cell wall elongation machinery. Nygaard et al. resolved the three-dimensional structure of this complex using cryo-electron microscopy (PDB ID: 8TJ3, resolution 3.0 Å) [3], revealing its conformational features within lipid nanodiscs and identifying two critical Lipid II binding cavities (A and B), which serve as donor and acceptor sites during peptidoglycan polymerization. Based on this structural information, we selected the RodA–PBP2 complex as the docking target to investigate the interaction mechanism of antimicrobial peptides with Gram-positive bacteria.

Molecular docking was performed using the LightDock online server (https://server.lightdock.org/). This platform accounts for the flexibility of both the ligand and the receptor, thereby approximating dynamic interactions under physiological conditions.

The specific workflow is as follows:

• Input Preparation: The three-dimensional structure of the antimicrobial peptide (ligand) was generated by AlphaFold3. Target protein structures were obtained from the PDB database. Missing regions, if any, were completed using homology modeling. Structures were then standardized, including removal of water molecules, addition of hydrogen atoms, and protonation optimization via molecular mechanics, to ensure the system approximates physiological conditions.
• Docking Setup: In the LightDock fast docking workflow, the flexible backbone mode was selected, allowing conformational adjustments of both the ligand and receptor backbone during docking, thereby more realistically simulating the binding state.
• Result Generation: LightDock samples potential binding sites of the ligand on the receptor surface using a swarm optimization algorithm, generating multiple candidate conformations. These are then scored and ranked using the platform’s built-in scoring function. The top-ranked binding modes were selected for subsequent analyses.

Binding Free Energy Analysis Based on PDBePISA

Binding free energy is a key indicator for evaluating the interaction strength between antimicrobial peptides and their receptors. We performed energy assessment of the docking results using the PDBePISA online tool (https://www.ebi.ac.uk/msd-srv/prot_int/). This platform predicts interaction energies and interface characteristics based on the three-dimensional structures of protein complexes.

Specifically, the top-ranked binding conformation from LightDock (in PDB format) was used as input. The server automatically identifies the ligand–receptor interface and calculates the free energy change (ΔG) along with associated interface parameters. A larger absolute value of the binding free energy indicates a more stable interaction between the antimicrobial peptide and its receptor. By further analyzing the energy components and the contributions of key residues, we can identify amino acid residues that play central roles at the binding interface, thereby elucidating the molecular basis of antimicrobial peptide activity.

By integrating the spatial binding modes obtained from LightDock with the energy analysis results from PDBePISA, we are able to systematically interpret the interaction mechanism of Ulink-AMP and its mutants with the RodA–PBP2 complex, providing a theoretical basis for subsequent structure-guided antimicrobial peptide engineering and functional optimization.

Results and Analysis

Flexible molecular docking of the selected antimicrobial peptide with the E. coli RodA–PBP2 complex (PDB ID: 8TJ3) using LightDock yielded the top ten potential binding conformations. Based on the predicted binding affinity from the docking scoring function, the lowest-energy binding mode was selected (Figure 1).

The results indicate that the antimicrobial peptide primarily occupies Lipid II binding cavity A of the RodA–PBP2 complex. The interaction interface spans both the transmembrane helical region of RodA and the periphery of the PBP2 catalytic domain.

This binding pattern suggests that the antimicrobial peptide may simultaneously target the GT and TP functional regions, thereby potentially interfering with peptidoglycan biosynthesis.

From the perspective of the three-dimensional binding mode, the antimicrobial peptide adopts an extended conformation that inserts into binding cavity A. The N-terminal residues of the peptide, such as Lys and Arg, form stable van der Waals interactions with RodA.

This binding pattern partially overlaps with the native binding mode of Lipid II at this site, suggesting that the antimicrobial peptide may competitively occupy binding cavity A, thereby interfering with the interaction between Lipid II and the RodA–PBP2 complex, ultimately affecting normal peptidoglycan biosynthesis.

The top-ranked binding conformation from LightDock was submitted to the PDBePISA server for analysis of binding free energy and interface characteristics. The results indicate that the primary binding interface between the antimicrobial peptide and the RodA–PBP2 complex has a contact area of approximately 850 Ų, with a predicted binding free energy (ΔG) of –15.2 kcal/mol, suggesting a relatively stable interaction between the two.

This value falls within the typical energy range for protein–ligand interactions, further supporting the potential of the antimicrobial peptide to effectively bind the complex at the structural level and exert its function (Figure 3).

Mutational Engineering of Antimicrobial Peptide

Rational Design

Molecular docking analysis using LightDock indicated that Ulink–AMP primarily interacts with the PBP2 subunit of the E. coli RodA–PBP2 complex through its second α–helical region. To validate and optimize this interaction, the PDBePISA server was employed to calculate the binding free energy (ΔG) and interface properties between Ulink–AMP and PBP2. The results demonstrated the formation of a stable complex, confirming the potential of this binding mode.

Based on this finding, we applied rational design by introducing single–point or multi–point mutations to enhance the binding affinity of the antimicrobial peptide for PBP2, thereby potentially improving its antimicrobial activity. We focused on analyzing amino acid residues at the binding interface and selected two sites located in key interaction regions for mutation: Glycine (G) at position 40 and Alanine (A) at position 42.

Mutation Strategy

Based on the analysis of the binding interface between Ulink–AMP and PBP2, we proposed the following mutation strategies:

• A42F mutation: Alanine (A) is a small hydrophobic residue, and in the docking model, the A42 residue points directly toward a hydrophobic pocket on the PBP2 surface. To enhance hydrophobic interactions at this site, we substituted it with Phenylalanine (F). The aromatic ring of phenylalanine not only provides more extensive hydrophobic contacts with the pocket but may also form π–π stacking interactions with aromatic residues on PBP2 (e.g., Tyr, Trp, Phe), introducing additional specific interactions and thereby significantly improving binding strength and specificity.
• G40A mutation: Glycine (G) lacks a side chain, resulting in high conformational flexibility, which can destabilize α–helices. We hypothesized that substituting the residue at position 40 with Alanine (A) would introduce a methyl side chain, reducing backbone conformational entropy and enhancing the stability of the α–helix in this region. A more stable helix can better insert into the binding pocket of PBP2, minimizing conformational rearrangement energy during binding and improving overall affinity.
• G40A/A42F double mutation: To explore the synergistic effect of the two optimization strategies, we designed a double mutant (116–a42f–g40a). We hypothesize that the helix stabilization conferred by the G40A mutation provides a more rigid structural framework for the A42F mutation, allowing the phenyl side chain to insert more precisely into the PBP2 target site, potentially achieving a 1 + 1 > 2 enhancement effect.

Computational Validation

We performed molecular docking of the three mutant variants — AMP–TB1 (A42F), AMP–TB2 (G40A), and AMP–TB3 (A42F & G40A) — with the RodA–PBP2 complex, followed by analysis of the binding free energy (ΔG) and interface properties of the top–ranked docking conformations using the PDBePISA server.

The computational results are summarized in the following table:

Based on the docking and energy analysis data, we drew the following conclusions:

• Overall trend: All designed mutant variants computationally exhibited stronger binding affinity compared to the wild-type peptide.
• A42F single mutation: The binding free energy decreased by 2.9 kcal/mol, accompanied by a significant increase in interface area. This indicates that enhancing hydrophobic interactions at key residues and introducing aromatic ring stacking is an effective strategy to improve binding strength.
• G40A single mutation: The binding free energy decreased by 1.3 kcal/mol, validating our hypothesis that stabilizing the α–helix scaffold can enhance binding affinity.
• G40A/A42F double mutation: Exhibited the strongest binding capability, with a binding free energy of –19.6 kcal/mol, which is lower than either single mutant. The effect exceeds the sum of the individual contributions, indicating a clear synergistic enhancement. Structural analysis further revealed that the helix stabilization conferred by G40A allows the peptide backbone to pre–organize into a conformation complementary to the PBP2 surface, enabling the phenyl side chain of A42F to insert more deeply and precisely into the hydrophobic pocket, thereby forming a tighter interaction network.

In summary, the simulation results indicate that rational introduction of the G40A and A42F mutations can significantly enhance the binding affinity of the antimicrobial peptide toward the PBP2 subunit, with the double mutant exhibiting the optimal binding potential, providing a clear direction for subsequent experimental optimization.

Experimental Validation

See the Engineering Success section for details.

To validate the modeling results, we designed and constructed three mutant variants — AMP–TB1 (A42F), AMP–TB2 (G40A), and AMP–TB3 (G40A & A42F) — based on pET–28a–eel–Ulink–AMP. Their antibacterial activities and minimum inhibitory concentrations (MICs) were subsequently evaluated.

We selected Escherichia coli BL21(DE3) and Bacillus subtilis ATCC 6633 as representative standard strains for testing, corresponding to typical Gram–negative and Gram–positive bacteria, respectively. This selection allowed us to comprehensively assess the broad–spectrum antibacterial potential of the peptides.

The results showed that all peptide samples, including the wild–type Ulink–AMP, exhibited significant antibacterial activity against both strains. Among the mutants, AMP–TB2 displayed the lowest MIC values, demonstrating superior inhibition of E. coli and B. subtilis compared to the wild–type sequence and the other mutants. This indicates that AMP–TB2 not only achieved the highest expression and purification efficiency, but also exhibited the greatest antibacterial potential in practical assays.

To enable a more intuitive comparison of the antibacterial efficacy among different peptides, an agar diffusion inhibition assay was performed. The diameter of the inhibition zones was measured to directly reflect each peptide’s inhibitory capacity on solid media, offering solid evidence for identifying the most promising candidate.

The next-day observations showed that all peptides at a concentration of 150 mg/mL formed clear inhibition zones against both strains, indicating antibacterial activity against both Gram–negative and Gram–positive bacteria. Among them, AMP–TB2 exhibited the strongest inhibition, producing an inhibition zone of 13.12 ± 0.45 mm against Bacillus subtilis, which was markedly superior to the wild–type peptide and the other mutants.

Combined with the MIC measurements and plate assay results, AMP–TB2 was further confirmed as the optimal candidate for subsequent studies.

Mechanism Explanation

By integrating modeling predictions and experimental data, it is evident that AMP–TB2 (G40A) exhibited the best performance among all designed mutants. MIC measurements showed that TB2 had the lowest minimum inhibitory concentrations against both E. coli and B. subtilis, and the inhibition zone assays also demonstrated the largest inhibition diameters. This functional advantage can be rationalized based on its structural features.

In the molecular docking and energy analyses, G40 was identified as a key residue within the second α–helix of the antimicrobial peptide. In the wild–type peptide, glycine (G), lacking a side chain, typically imparts high backbone flexibility, but at the cost of reduced α–helical stability. Mutation to alanine (A) introduces a methyl side chain, which restricts the conformational freedom of the backbone, thereby significantly enhancing the local α–helix stability. This stabilization reduces the energetic cost of backbone rearrangement during binding, enabling the peptide to adopt a more favorable pre–organized conformation for insertion into the RodA–PBP2 binding pocket.

Further interface energy calculations corroborated this inference: the G40A mutant exhibited a lower binding free energy and a larger interface area compared to the wild–type, indicating a more stable peptide–RodA–PBP2 complex. This optimized binding mode directly translated into the observed functional performance, with AMP–TB2 displaying significantly enhanced antibacterial activity against both representative bacterial strains compared to the wild–type peptide and other mutants.

Compared to AMP–TB2 (G40A), the other two mutants did not exhibit superior performance.

AMP–TB1 (A42F): In the docking model, residue A42 faces a hydrophobic pocket on the PBP2 surface. Initially, we hypothesized that replacing the small alanine with phenylalanine, which carries an aromatic ring, could enhance hydrophobic interactions and potentially introduce π–π stacking. However, experimental results indicated that AMP–TB1 did not achieve the expected antibacterial activity, with MIC values noticeably higher than those of TB2. This discrepancy can be rationalized by the modeling analysis: although phenylalanine provides potential hydrophobic and aromatic interactions, its larger size may introduce steric hindrance within the confined binding pocket, reducing the conformational adaptability of the peptide and consequently diminishing overall binding efficiency. This suggests that enhancing hydrophobic interactions does not always confer positive effects, and the spatial accommodation of the binding pocket is an equally critical constraint.

AMP–TB3 (G40A/A42F double mutant): We initially hypothesized that the stabilizing effect of G40A could provide a more favorable structural framework for the phenyl ring of A42F, potentially resulting in a synergistic effect. However, experimental results showed that TB3 did not surpass TB2 in antibacterial activity. Docking analyses suggested that although the double mutation indeed stabilized the α–helix, the steric hindrance introduced by A42F persisted, which may have offset some of the advantages conferred by G40A. In other words, the double mutation did not exhibit a “1 + 1 > 2” effect, and due to structural accommodation constraints, its binding and functional performance remained inferior to that of AMP–TB2.

In summary, the modeling and experimental results consistently indicate that the G40A mutation (AMP–TB2) represents the optimal modification strategy. Its advantage arises from the stabilization of the local α–helix, allowing the antimicrobial peptide to adopt a more favorable pre–organized conformation with lower rearrangement energy for insertion into the binding pocket, thereby forming a tighter and more stable interface interaction. This structural optimization is directly reflected in the experimental data as the lowest MIC values and the largest inhibition zones. In contrast, the A42F single mutation exhibited reduced binding efficiency due to steric hindrance, and the double mutant failed to achieve synergistic enhancement, highlighting the need to balance spatial accommodation with energetic optimization. This study not only clarifies the mechanistic basis for the superior performance of TB2 but also validates the guiding value of modeling in rational design of antimicrobial peptides, providing a solid foundation for subsequent structural optimization and functional expansion.

Part2：Verification of the Broad-Spectrum Antibacterial Potential of AMP-TB2

Methods

Target Protein Selection

To further investigate the mechanism of action of the antimicrobial peptide against other common human–animal zoonotic pathogens, we selected the following proteins as docking targets: penicillin-binding protein 2 (PBP2, PDB ID: 3DWK) from Staphylococcus aureus, glycosyltransferase (GT, PDB ID: 5N80) from Salmonella enterica, and N-acetyltransferase PseH (PDB ID: 4XPK) from Campylobacter jejuni.

Staphylococcus aureus is a common human–animal zoonotic pathogen that can cause skin and soft tissue infections, sepsis, and drug-resistant nosocomial infections, posing a significant threat to public health. Its pathogenicity relies on a robust cell wall structure, and PBP2 serves as a critical transpeptidase in cell wall biosynthesis. The enzymatic activity of PBP2 directly determines the cross-linking stability of the bacterial peptidoglycan network, making it a key target for multiple classes of antibiotics. By performing molecular docking of the high-confidence structural conformation of AMP–TB2 with this receptor, we can predict whether the antimicrobial peptide can stably bind to its catalytic region, thereby inferring its potential inhibitory mechanism against S. aureus [4].

Salmonella enterica is a common human–animal zoonotic pathogen that can cause gastroenteritis and, in severe cases, systemic infections. Its pathogenicity relies on the integrity of the cell wall and outer membrane polysaccharides. Glycosyltransferases (GTs) play a key role in the biosynthesis of lipopolysaccharides (LPS) and other glycan structures, serving as essential factors for maintaining the outer membrane barrier and resisting host immune defenses. Therefore, GTs are not only critical for bacterial structural stability but are also considered potential targets influencing Salmonella virulence and survival [5]. By performing molecular docking targeting GT, we aim to evaluate whether AMP–TB2 can stably interact within the enzyme’s active site, thereby inferring its potential to disrupt LPS biosynthesis and outer membrane function.

Campylobacter jejuni is a major causative agent of acute gastroenteritis and peripheral neuropathies in humans, such as Guillain–Barré syndrome. Its pathogenicity heavily depends on the motility of its flagella. Literature reports indicate that PseH catalyzes the third step in the flagellin glycosylation pathway, which is essential for flagellar assembly and normal motility. Therefore, PseH plays a central role in both the growth and pathogenic processes of C. jejuni, and its structural and functional features make it a rational target for therapeutic intervention. By performing molecular docking analysis targeting PseH, we can evaluate whether the antimicrobial peptide can interfere with flagellar formation and function. Notably, this mode of action aligns more closely with an anti-virulence strategy, aiming not to directly kill bacteria but to reduce pathogenicity by impairing motility and colonization, thereby exerting inhibitory effects [6].

Molecular Docking Based on AutoDock Vina

In the molecular docking experiments, the receptor protein structures were first preprocessed using PyMOL. Co-crystallized ligands and water molecules were removed, and the protein conformations were optimized by assigning charges, adding hydrogen atoms, and merging non-polar hydrogens. Subsequently, AutoDockTools 1.5.7 was employed to add polar hydrogens and assign charges to the receptor, enabling accurate energy calculations.

Docking simulations were performed using AutoDock Vina, whose empirical scoring function integrates van der Waals interactions, Coulombic forces, hydrogen bonding, and hydrophobic effects. The scoring function is minimized to identify the most stable and probable binding modes. The ligand was the three-dimensional structure of AMP-TB2 predicted by AlphaFold 3, while receptor protein structures were obtained from the PDB database.

During the simulation, AutoDock Vina performed parameter sampling and conformational search using a fast and efficient annealing algorithm, generating multiple potential docking poses. Ultimately, the model with the lowest binding energy was selected as the optimal binding mode, and the docking conformation was visualized and analyzed for interface features using PyMOL.

Results and Analysis

Overall, AMP-TB2 inserts into the vicinity of the penicillin-binding site 2 of Staphylococcus aureus PBP2 in an extended conformation, forming a large interface with the receptor. A closer view reveals that its hydrophobic residues engage in stacking interactions with hydrophobic amino acids within the binding pocket, while certain polar residues establish hydrogen bonds with key sites, stabilizing the complex. This binding mode suggests that AMP-TB2 may competitively occupy the catalytic region of PBP2, hindering its normal transpeptidase activity and thereby interfering with the cross-linking of peptidoglycan in the cell wall, providing structural evidence for its antibacterial mechanism in Gram-positive bacteria.

The optimal docking conformation of AMP-TB2 with the glycosyltransferase from S. enterica (PDB ID: 5N80) shows that the peptide (yellow) adopts a curved, extended conformation along the receptor surface, partially penetrating its active pocket. Key residues of the peptide form stable interactions with amino acids at the receptor’s active center, including both hydrophobic contacts and specific hydrogen-bond-mediated interactions. Notably, AMP-TB2 occupies a position in the binding pocket that significantly overlaps with the substrate binding site, suggesting that it may competitively block the natural substrate from entering, thereby impairing the glycosyltransferase’s catalytic function in lipopolysaccharide synthesis.

In the docking with Campylobacter jejuni N-acetyltransferase PseH (PDB ID: 4XPK), AMP-TB2 (yellow) adopts a curved helical conformation spanning the surface of PseH and forms tight contacts with its active pocket region. Multiple residues of the antimicrobial peptide penetrate the binding cavity, establishing hydrogen bonds and hydrophobic stacking interactions with key amino acids near the catalytic center (highlighted in purple), thereby stabilizing the binding mode. The binding site significantly overlaps with the substrate-binding region, suggesting that AMP-TB2 may competitively occupy the substrate site, blocking PseH’s catalytic activity in the flagellar glycosylation pathway. This mechanism implies that the peptide could weaken the formation and function of flagella in C. jejuni, thereby reducing its motility and colonization ability within the host.

Although AMP-TB2 exhibits competitive binding characteristics across all three pathogen targets, the underlying mechanisms differ substantially. In S. aureus PBP2, AMP-TB2 embeds into the penicillin-binding site, inhibiting transpeptidase activity and directly interfering with peptidoglycan cross-linking—a classic cell wall synthesis inhibition mode. In S. enterica glycosyltransferase, AMP-TB2 occupies the active pocket required for LPS synthesis, likely blocking substrate entry and weakening outer membrane stability, reflecting a cell barrier disruption mechanism. In contrast, in C. jejuni PseH, AMP-TB2 functions via an anti-virulence strategy: by covering the substrate-binding site, it inhibits flagellar glycosylation, thereby impairing flagella assembly and motility, reducing bacterial migration and colonization rather than directly killing cells.

In summary, these docking results collectively demonstrate that AMP-TB2 can form stable interactions with multiple key targets and impair pathogen growth or virulence through distinct mechanisms, providing structural evidence for its broad-spectrum antibacterial activity.

Conclusion

Through structural prediction and molecular docking, we identified key mutation sites such as G40A and A42F, and energy/interface analyses suggested that the G40A mutation (AMP-TB2) can significantly enhance α-helix stability and binding affinity. Experimental validation further confirmed that TB2 exhibits the lowest MIC and strongest antibacterial activity against E. coli and B. subtilis, making it the optimal candidate. Building on this, we extended the analysis to its binding modes with S. aureus PBP2, S. enterica GT, and C. jejuni PseH. The results show that TB2 can competitively interact with multiple targets via distinct mechanisms, explaining its broad-spectrum antibacterial potential. These findings not only reveal the molecular basis for TB2’s superiority but also highlight the guiding value of computational modeling in the rational design and experimental validation of antimicrobial peptides.

Reference

• [1]
  Krissinel, E., & Henrick, K. (2007). Inference of macromolecular assemblies from crystalline state. Journal of Molecular Biology, 372(3), 774–797. Protein interfaces, surfaces and assemblies’ service PISA at the European Bioinformatics Institute.
• [2]
  Jiménez-García, B., Roel-Touris, J., Romero-Durana, M., Vidal, M., Jiménez-González, D., & Fernández-Recio, J. (2018). LightDock: A new multi-scale approach to protein–protein docking. Bioinformatics, 34(1), 49–55.
• [3]
  Nygaard, R., Graham, G. C. L. B., Dufrisne, M. B., Colburn, J. D., Pepe, J., Hydorn, M. A., et al. (2023). Structural basis of peptidoglycan synthesis by E. coli RodA–PBP2 complex. Nature Communications, 14(1), 5151.
• [4]
  Łęski, T., & Tomasz, A. (2005). Role of Penicillin-Binding Protein 2 (PBP2) in the antibiotic susceptibility and cell wall cross-linking of Staphylococcus aureus. Journal of Bacteriology, 187(5), 1815–1824.
• [5]
  Bravo, D., Silva, C., Carter, J. A., Hoare, A., Alvarez, S. A., Blondel, C. J., … & Contreras, I. (2008). Lipopolysaccharide O-antigen is required for Salmonella enterica serovar Enteritidis virulence in mice. FEMS Immunology & Medical Microbiology, 52(3), 267–275.
• [6]
  van Alphen, L. B., Wenzel, C. Q., Richards, M. R., Fodor, C., Ashmus, R. A., Stahl, M., … & Szymanski, C. M. (2015). Structural and functional characterization of PseH, an N-acetyltransferase involved in pseudaminic acid biosynthesis in Campylobacter jejuni. Biochimica et Biophysica Acta (BBA) - General Subjects, 1850(9), 1661–1672.