Controlling Nanonet Morphology via Residue‐Specific Modulation of β‐Hairpin Peptide for Enhanced Bacterial Trapping

Abstract

Precise control over peptide nanonet architecture is instrumental in advancing the development of antibacterial nanonets. Here, a novel design strategy is presented to control bacteria nanonet morphology through rational modification of the β‐hairpin side strands, leveraging the unique chemical properties of amino acid side chains. By fine‐tuning both the termini and aromaticity of the hydrophobic residue, the W‐W₁₃ peptide is engineered to form increased nanofibers interweaving on bacterial surfaces, forming a tightly interwoven nanonet that effectively traps and kills both E. coli and S. aureus. In contrast, asymmetric glutamic acid substitutions on the cationic residues of the E‐E₁₃ ^ASYM peptide redirect the nanofibers to self‐interweave, forming extensive nanonets with minimal bacterial coverage and no antibacterial activity. Using these nanonets with distinct morphologies and function, it is demonstrated that the formation of tightly interwoven nanonets on bacterial surfaces significantly reduces the spread of motile E. coli and P. aeruginosa, outperforming both loosely trapping nanonets and conventional potent antibiotics. The findings pave the way for the development of novel peptide‐based nanonets, offering a promising strategy to target bacterial motility and prevent spreading of bacteria.

From Sequence to Function: Modifying the side strands of β‐hairpin peptides with aromatic or negatively charged residues enables the tuning of their self‐assembly to create bacteria‐specific nanonets with distinct morphologies and functions. Tight nanonets effectively encage and kill motile bacteria to reduce post‐entrapment escape and spread, while loose nanonets without antibacterial activity fail to contain bacterial outgrowth.

1. Introduction

Bacterial trapping offers a promising strategy to combat antibiotic resistance by targeting mechanisms that help bacteria evade antibiotics, such as internalization in host cells or escaping areas with high local drug concentrations.^{[ 1} , ² , ³ , ⁴ , ⁵ , ^{6 ]} Human‐defensin 6 (HD6) peptide forms an effective bacteria‐trapping nanonet in the human small intestine,^{[ 7} , ^{8 ]} but recreating such nanonet using short synthetic peptides is challenging due to the limited understanding of the design principles. Self‐assembling hybrid peptides can form nanofibers that trap bacteria,^{[ 9} , ¹⁰ , ^{11 ]} but these nanofibers lacked antibacterial activity, risking bacterial outgrowth and spread. Furthermore, the spontaneous self‐assembly of these peptides may lack spatiotemporal control, limiting their applicability for systemic use and making them more suitable for localized treatments such as topical application on the skin. Additionally, uncontrolled assembly can compromise long‐term stability, affecting both storage and ease of administration.^{[ 12 ]} Although HD6‐like nanonets are not naturally found on the human skin, synthetic peptide nanonets represent a novel strategy for treating chronic skin infections and inflammation. These nanonets can be applied topically to trap‐and‐kill bacteria at the infection site, thereby preventing bacterial dissemination into the bloodstream. However, to fully realize this potential, improved peptide designs are still needed to enhance bacterial responsiveness and selectivity while maintaining biofunctions.

β‐hairpin peptides, with controllable conformational bias and self‐assembly,^{[ 13} , ^{14 ]} have been shown to form cytocompatible and antibacterial peptide hydrogels.^{[ 15} , ^{16 ]} These peptides, featuring (XY)_n repeats of hydrophobic and cationic residues flanking a central turn motif,^{[ 17 ]} fold into β‐hairpin facial amphiphiles with distinct hydrophobic and cationic regions, which is a defining structural characteristic of antimicrobial peptides (AMPs).^{[ 18 ]} In our previous work, we designed a series of β‐hairpin peptides with novel turn motifs that exhibited potent antibacterial activity.^{[ 19 ]} Introducing the novel turn motif (Leu‐Thr‐Ala) in the 15‐mer BTT1‐3A enabled it to self‐assemble into nanonets specifically in the presence of bacteria, rather than forming non‐specific peptide hydrogels.^{[ 20 ]} Initially, BTT1‐3A was a highly water‐soluble random coil. Upon binding to bacterial pathogen‐associated molecular patterns (PAMPs) such as lipopolysaccharide (LPS) or lipoteichoic acid (LTA), it exhibited conformational bias by folding into β‐hairpin and self‐assembling into nanofibers, leading to the formation of bacteria‐trapping nanonets.^{[ 20} , ^{21 ]} Furthermore, BTT1‐3A demonstrated broad‐spectrum antibacterial activity, forming extensive nanonets that trap‐and‐kill various bacterial species while also sequestering endotoxins to reduce inflammation.^{[ 20} , ^{22 ]} While BTT1‐3A nanofibers form extensive nanonets, they tend to entangle with themselves rather than tightly wrapping around bacterial surfaces. This represents a limitation of the current system. Although PAMPs‐induced fibrillation enables improved temporal control, the resulting nanofibers often fail to tightly wrap around bacterial surfaces. Such loose entrapment may allow bacteria to undergo post‐trapping escape and dissemination. This limits the effectiveness of peptide nanonets in controlling infection, highlighting the need for further optimization of peptide design to fine‐tune both peptide–bacteria and peptide–peptide interactions for improved bacterial trapping. We hypothesize that improving the ability of peptide nanofibers to tightly interweave on bacterial surfaces will enhance their effectiveness in trapping and killing bacteria.

Inspired by spiders that weave extensive nets to initially trap prey and then tightly interweave them to confine prey and prevent escape, our next design goal is to achieve better control of peptide nanofibers interweaving on bacterial surfaces to enhance trapping and limit bacteria spread. By strategic substituting residues on the side strands with either aromatic or negatively charged residues, we created nanonets with distinct morphologies: either tightly interwoven nanonets on bacterial surfaces or extensive hydrogel‐like nanonets with minimal bacterial coverage. Furthermore, we demonstrated that these tightly interwoven nanonets effectively encage motile bacteria, significantly reducing their spread compared to loosely trapping nanonets.

2. Results and Discussions

2.1. Design Rationale for Tuning Nanonet Morphology and Activity

To this end, we designed a library of BTT1‐3A analogues which retained the turn motif (Leu‐Thr‐Ala) but were systematically modified at side strands via three major approaches to control bacteria nanonet morphology while preserving bacterial specificity: 1) shortening the peptides from both termini, 2) fine‐tuning aromaticity or 3) reducing the overall positive charge (Figure  1 ).

Figure 1.

Schematic illustration showing the rational design of β‐hairpin peptide to form bacteria nanonet. BTT1‐3A fold into β‐hairpin after binding to bacterial LPS or LTA. The β‐hairpin peptide stacks into nanofibrils to form nanonet and trap bacteria. Three modification strategies were applied to the β‐hairpin side strands, focusing on shortening of the peptide, fine‐tuning aromaticity, or reducing overall positive charge. Key residues were highlighted in red.

First, the peptides were shortened from both ends to identify the minimum side strand length required for fibrillation. Next, to create tightly interwoven nanonets on bacteria surfaces for trapping and killing them, hydrophobic residues at both cross ends of the β‐hairpin peptides were symmetrically substituted with phenylalanine or tryptophan, known to enhance bacterial membrane interaction and bactericidal activity of typical AMPs.^{[ 23} , ²⁴ , ²⁵ , ^{26 ]} Crucially, the overall hydrophobicity of the peptides was carefully modulated to stabilize the β‐hairpin and promote nanofiber fibrillation through intramolecular and intermolecular π‐π stacking, while preventing spontaneous self‐assembly into peptide hydrogels. Seven peptides were designed with varying combinations of length and symmetrical cross‐end aromatic pair motifs (F‐F, diF‐F or W‐W) to determine the optimal configuration (Table  1 ). Conversely, to demonstrate that peptide nanonets without antibacterial activity can trap bacteria but are less effective at reducing bacteria spread, we designed nanofibers that self‐interweave to form extensive nanonets rather than interweaving on the bacterial surface. This was achieved by substituting lysine with glutamic acid at both ends to reduce the net positive charge. This modification was devised to stabilize the β‐hairpin and enhance nanofiber fibrillation through electrostatic attractions. However, careful placement of glutamic acids may be important to avoid destabilizing the β‐hairpin due to negative charge repulsion; therefore, two peptides with different glutamic acid positions were designed. Additional modifications included substituting leucine with valine to reduce hydrophobicity while retaining high β‐sheet‐forming propensity.^{[ 27 ]} Leucine with flexible side chain was substituted with rigid isoleucine to prevent α‐helix formation and promote β‐sheet formation.^{[ 28 ]} Some lysine residues were replaced with arginine to enhance antimicrobial activity.^{[ 29 ]}

Table 1.

Peptide designs and their physicochemical parameters. All peptides were amidated at C‐terminus. The turn motif was underlined, and key residues are bold. Subscript numbers denote the length of the peptides.

Peptide	Sequence	Length	Charge a )	Average hydrophilicity a )
BTT1‐3A	LKLKLKLTAKLKLKL	15	7	0.30
1. Reduce length
btt1‐3A13	KLKLKLTAKLKLK	13	7	0.62
btt1‐3A11	LKLKLTAKLKL	11	5	0.19
2. Increase aromaticity
F‐F15	LKFRVKLTAKVRFKL	15	7	0.25
diF‐F15	LKFRFKLTAKFRFKL	15	7	0.11
F‐F13	KFRVKLTAKVRFK	13	7	0.56
diF‐F13	KFRFKLTAKFRFK	13	7	0.41
W‐W13	KWRVKLTAKVRWK	13	7	0.42
F‐F11	FRVKLTAKVRF	11	5	0.12
W‐W11	WRVKLTAKVRW	11	5	−0.05
3. Reduce positive charge
E‐E13	EVRIKLTAKIRVE	13	3	0.67
E‐E13 ASYM	EVRIKLTAKIEVR	13	3	0.67

^a) The charge and average hydrophilicity of peptide was calculated using Bachem peptide calculator. (https://www.bachem.com/knowledge‐center/peptide‐calculator/)

2.2. Bacteria‐Responsive Self‐Assembly of Peptides

Ab initio secondary structure prediction using D‐I‐TASSER indicated that all peptides were likely to adopt extended conformations with a high propensity to fold into β‐hairpins (Figure S1, Supporting Information).^{[ 30 ]} Circular dichroism (CD) spectroscopy was performed to examine the peptide secondary structures. All peptides remained random‐coiled in phosphate‐buffered saline (PBS) (Figure  2A) and folded into β‐sheets upon binding to bacterial PAMPs (Figure 2B,C). The only exception was diF‐F₁₅, which formed β‐sheets in PBS alone, suggesting an inability to respond to bacterial PAMPs. We then assessed the extent of amyloid fibrillation by measuring the fluorescence of amyloid‐specific K114 dye in the presence of bacterial PAMPs. Specifically, BTT1‐3A, F‐F_15, F‐F₁₃, diF‐F₁₃, W‐W₁₃ and E‐E₁₃ ^ASYM fibrillated rapidly within 30 minutes upon binding to bacterial PAMPs, showing significantly higher K114 signals compared to the non‐fibrillating BTT4 baseline control (Figure 2D,E). In contrast, diF‐F₁₅ spontaneously fibrillated in PBS regardless of PAMPs’ presence (Figure S2A, Supporting Information), indicating a lack of specificity. These findings show that the six peptides are bacteria‐responsive, folding into β‐sheet structures in the presence of bacterial PAMPs, which in turn triggers nanofiber formation. Unlike peptides that self‐assemble non‐specifically in buffer (e.g., diF‐F₁₅), these peptides undergo PAMPs‐induced fibrillation, enabling better temporal control over nanonet formation and distinguishing them from conventional peptide hydrogels. To define the minimal sequence requirements for this responsive fibrillation, we systematically truncated BTT1‐3A. Removing two terminal leucine residues in BTT1‐3A significantly decreased fibrillation. Further shortening to 11‐mer peptide abolished fibrillation, as indicated by the negligible fluorescence signals of btt1‐3A₁₁. Since incorporating F‐F or W‐W motifs into the 11‐mer peptides did not enhance fibrillation, we focused on the 13‐ and 15‐mer peptides to study the impact of different side strand modifications on controlling the peptide fibrillation.

Figure 2.

Secondary structures and fibrillation capability of the peptides. A–C) CD spectra of 50 µm peptides were measured with (A) PBS alone, (B) 0.2 mg mL⁻¹ LPS, and (C) 0.2 mg mL⁻¹ LTA. The blue dashed line indicates β‐sheet structure of diF‐F₁₅ in PBS. D–E) Percentage increase in K114 fluorescence signal for each 20 µm peptide after incubation for 0.5, 1, and 2 hours in the presence of 5 µg mL⁻¹ (D) LPS and (E) LTA. The grey dashed line represents the baseline for non‐fibrillating control peptide (BTT4). Experiments were repeated three times, with data expressed as mean ± SE. Statistical significance compared to the non‐fibrillating control BTT4 was evaluated using Dunnett's multiple comparisons (ns: p > 0.05, ****: p ≤ 0.0001).

Next, we showed that F‐F₁₃, diF‐F₁₃, and W‐W₁₃, with symmetrical substitutions of F‐F, diF‐F, and W‐W motifs, respectively, enhanced the fibrillation of the 13‐mer peptide significantly. Interestingly, the 15‐mer peptide F‐F₁₅, with a F‐F motif, showed lower fibrillation compared to BTT1‐3A. However, the additional phenylalanine pairs in diF‐F₁₅ significantly increased the K114 fluorescence signal, even in PBS alone (Figure S2A, Supporting Information). Using confocal and scanning electron microscopy (SEM), we confirmed the spontaneous and uncontrolled self‐assembly of diF‐F₁₅ into a peptide hydrogel (Figure S2B,C, Supporting Information). Since diF‐F₁₃, which had the same sequence as diF‐F₁₅ but lacked the terminal leucine, did not form β‐sheet in PBS alone (Figure 2A), our findings suggest that fine‐tuning both the terminal end and aromaticity of the hydrophobic residue of the 13‐mer β‐hairpin peptide was crucial to enhance nanofibers fibrillation while maintaining bacterial responsiveness, thereby avoiding spontaneous self‐assembly into peptide hydrogels in the absence of bacterial triggers. Moreover, we showed that asymmetrical substitution of glutamic acids at both ends of the 13‐mer E‐E₁₃ ^ASYM improved fibrillation, while E‐E₁₃ with symmetrical placement did not. This observation supports our hypothesis that optimal positioning of glutamic acids was crucial to promote fibrillation through electrostatic attraction while minimizing negative charge repulsion. Overall, CD spectra and K114 fibrillation assays suggest that the rational design of peptides capable of selectively undergoing conformational changes and fibrillation upon binding to bacterial PAMPs is crucial for developing bacteria‐responsive nanonets, rather than non‐specific peptide hydrogels.

2.3. Fine‐Tuning Peptide Motifs Controls Nanonet Architecture

To enable confident comparisons of nanonet morphology, we first optimized a high peptide‐to‐bacteria ratio to ensure sufficient bacterial density and consistent nanonet formation across imaging fields. SEM images revealed that the fibrillating peptides formed nanonets of distinct morphologies when incubated with the Gram‐negative E. coli and the Gram‐positive S. aureus (Figure  3A). W‐W₁₃ self‐assembled into thin nanofibers which densely interwove around individual bacteria, forming a tight nanonet with higher surface coverage on E. coli than S. aureus, possibly due to differences in bacterial membrane composition. For E‐E₁₃ ^ASYM, which is expected to interact minimally with bacterial membranes, we used a higher concentration (100 µm) to test whether excess peptide would preferentially self‐assemble into nanonets rather than bind to bacterial surfaces. The SEM images confirmed that the nanofibers of E‐E₁₃ ^ASYM interwove tightly among themselves to form extensive nanonets in the spaces between the bacteria, with minimal membrane contact and low surface coverage to loosely trap both E. coli and S. aureus. These findings support our design principle that E‐E₁₃ ^ASYM favors peptide–peptide self‐interaction over peptide–bacteria binding, enabling sequence‐based control of nanonet architecture.

Figure 3.

Residue‐specific modifications in β‐hairpin side strands resulted in nanonets with different morphologies. A) SEM images of untreated and treated E. coli ATCC 25 922 and S. aureus ATCC 29 737 with 40 µm of W‐W₁₃ or 100 µm of E‐E₁₃ ^ASYM for 2 hrs. The red arrows represent nanonet densely interweave around bacteria, while the yellow arrows represent nanonet extensively interweave in the space between the bacteria. Magnification = 15 000×. Scale bar = 1 µm. At least three independent fields were imaged per sample to ensure representativeness. B) Summary of peptide nanonet surface coverage and extensiveness. (Green: High, Yellow: Moderate, Red: Low).

Other fibrillating peptides formed nanonets with varying degrees of extensiveness and surface coverage against both bacteria strains (Figure 3B; Figures S3,S4, Supporting Information). The shortened peptide btt1‐3A₁₃ formed less extensive nanonets with low surface coverage around E. coli or random aggregated structures around S. aureus, consistent with its low K114 fluorescence signal (Figure 2D,E). We observed that BTT1‐3A and F‐F₁₅ nanofibers tended to aggregate and clump, creating extensive nanonets with moderate surface coverage that loosely trapped both E. coli and S. aureus. When the terminal leucine was removed, F‐F₁₃ nanofibers showed less aggregation and interwove extensively to loosely trap E. coli with moderate surface coverage. Conversely, F‐F₁₃ slightly increased surface coverage on S. aureus but with less extensive nanonets. The additional phenylalanine pairs in diF‐F₁₃ promoted tighter interweaving of nanofibers on the surface of E. coli, similar to W‐W₁₃ with high bacterial surface coverage. However, diF‐F₁₃ formed nanonets similar to F‐F₁₅ and BTT1‐3A against S. aureus but with less aggregated structures. Overall, these nanonets demonstrated the ability to agglutinate both E. coli and S. aureus with varying effectiveness (Figures S5,S6, Supporting Information).

Taken together, our findings suggest that to develop tightly interwoven nanonet on bacteria surfaces, removing terminal leucine may prevent undesirable nanofibers aggregation, potentially redirecting them to interweave on bacterial membrane. Notably, W‐W₁₃ and diF‐F₁₃, which have similar overall hydrophobicity (Table 1), exhibited increased nanofiber interweaving on E. coli membrane. However, on S. aureus, such tight nanonets were observed only for W‐W₁₃, likely due to the presence of the indole ring of tryptophan, which may enhance binding to the anionic lipoteichoic acid‐rich cell wall of Gram‐positive bacteria through additional hydrogen bonding and stronger π–cation interactions compared to phenylalanine.^{[ 23} , ²⁴ , ^{31 ]} Conversely, to create extensive hydrogel‐like nanonet with minimal bacterial coverage, asymmetrical substitution of lysine with glutamic acids can reduce the net positive charge of the peptide, enhancing self‐interweaving of nanofibers through electrostatic attraction while minimizing binding to the negatively charged LPS or LTA on bacterial membranes.

2.4. Peptide Design Modulates Nanonet Function in Bacterial Killing and Trapping

Next, we assessed the antibacterial activity of the newly designed peptide nanonets (Table S1 and Figure S7A, Supporting Information). These nanonets with different morphologies were able to trap‐and‐kill both E. coli and S. aureus with antibacterial activity comparable to the parent peptide BTT1‐3A, except E‐E₁₃ ^ASYM. Notably, W‐W₁₃ and F‐F₁₅ demonstrated a two‐fold improvement in antibacterial activity against both E. coli and S. aureus compared to the parent peptide BTT1‐3A. K114 binding assays revealed that both peptides exhibited lower fibrillation than BTT1‐3A upon binding to LPS and LTA, respectively (Figure 2D,E). This decrease in fibrillation may indicate that a larger proportion of the free peptides in solution remains available to directly interact with and disrupt bacterial membranes, rather than assembling into nanofibrils. This observation aligns with our previously proposed mechanism suggesting a trade‐off between peptide fibrillation and membrane‐disruptive antibacterial activity.^{[ 20 ]} A similar trend was observed for F‐F₁₃, which showed enhanced activity against E. coli—correlating with lower fibrillation in response to LPS—but not against S. aureus, where higher in fibrillation upon LTA binding was observed (Figures 2D,E and S7A, Supporting Information). These findings suggest that an optimal balance between self‐assembly and bacterial membrane interaction is critical for designing effective trap‐and‐kill nanonets. Further mechanistic studies are required to fully understand the various factors influencing peptide–bacteria interactions, self‐assembly, and antibacterial activity, which will support the development of more robust design principles for enhancing nanonet function. In contrast, E‐E₁₃ ^ASYM formed a trap‐only nanonet but lacked antibacterial activity. This is likely due to its significantly reduced binding to bacterial PAMPs (Figure S7B,C, Supporting Information), reinforcing the importance of bacterial PAMPs interaction in initiating both self‐assembly and antibacterial effects.

Additionally, removing the terminal leucine in the 13‐mer peptides improved cytocompatibility towards HaCaT keratinocytes compared to the parent peptide BTT1‐3A (Figure S8, Supporting Information). Specifically, the tight trap‐and‐kill peptide nanonet W‐W₁₃ exhibited a two‐fold improvement in CC₅₀ (69.3 µm) relative to BTT1‐3A (32.1 µm). Despite the modest improvement, W‐W₁₃ caused significantly less membrane damage and maintained normal HaCaT morphology, unlike BTT1‐3A, which led to severe cell membrane disruption (Figure S9, Supporting Information). Although W‐W₁₃ may have potential for topical application to locally trap‐and‐kill bacteria on the skin surface, its CC₅₀ value still presents a limitation for use at higher concentrations. Thus, further optimization of the peptide design is needed to enhance biocompatibility without compromising nanonet formation or antibacterial efficacy. However, as the exact design rules for forming tight trap‐and‐kill nanonets remain unclear, further modifications of the W‐W₁₃ motif using unnatural amino acids to fine‐tune the overall peptide physicochemical properties may represent a promising strategy to enhance bacterial selectivity and improve cytocompatibility while preserving tight nanonet formation as well as antibacterial activity.

Finally, we evaluated the effectiveness of nanonet with different morphologies in trapping and preventing bacterial spread. Since S. aureus is non‐motile, we speculated that our nanonets with improved antibacterial activity, such as W‐W₁₃ and F‐F₁₅, would effectively trap‐and‐kill the bacteria within the net, preventing their spread compared to existing nanofibers that lack bactericidal activity.^{[ 9} , ^{10 ]} Next, using soft‐agar motility assay that allows motile Gram‐negative bacteria like E. coli and P. aeruginosa to move freely and proliferate, we examined the trapping efficiency of different nanonets in preventing motile bacteria to escape from entrapment and spread. For this study, we selected W‐W₁₃ and diF‐F₁₃, which formed tight nanonets that encage and kill E. coli; BTT1‐3A, which formed loose nanonet and kill E. coli; and E‐E₁₃ ^ASYM, which formed loose nanonets without antibacterial activity. We first confirmed that these peptides formed nanonets with similar morphology and killing efficiency against P. aeruginosa as observed with E. coli (Figure S10 and Table S2, Supporting Information). Next, we optimized peptide concentrations to ensure sufficient bacterial viability to assess motility, while also accounting for differences in antibacterial activity. This allowed clear motility halo formation and evaluation of trapping ability.

In this assay, bacteria were first treated with peptides for only 30 minutes to allow nanonet formation. The nanonet‐trapped bacteria were then inoculated into soft‐agar plates to evaluate their ability to prevent bacterial escape, grow and spread after 14 hours. Motile E. coli and P. aeruginosa actively spread to form clear motility halos, and treatment with the potent antibiotic polymyxin B minimally reduced the diameter of motility halos (Figure S11, Supporting Information). This suggests that motile bacteria can escape from areas with high local drug concentrations, thereby avoiding death.^{[ 3} , ⁴ , ⁵ , ^{6 ]} When both bacteria were treated with equal concentrations of BTT1‐3A and diF‐F₁₃, which have similar antibacterial efficiency (Figure S7A and Table S2, Supporting Information), only the tight diF‐F₁₃ nanonet significantly reduced the motility halo and inhibited bacterial spreading, whereas the loose BTT1‐3A nanonet did not (Figure  4 ; Figure S11, Supporting Information). The tight W‐W₁₃ nanonet with improved antibacterial activity also effectively inhibited bacterial spread at lower concentrations. Our results clearly demonstrated that tightly interwoven nanonets on bacterial membrane was crucial for encaging and killing bacteria effectively, preventing their escape and spread. In contrast, E‐E₁₃ ^ASYM, which lacks antibacterial activity, was tested at a substantially higher concentration of 100 µm to evaluate whether extensive nanonets alone could prevent bacterial spread. The results showed that while the nanonets subtly restricted bacterial motility at this high concentration, the viable bacteria could still potentially outgrow and escape entrapment without effective killing.

Figure 4.

Tight trapping by peptide nanonets reduced the spread of motile bacteria. Images of soft agar plates A,B) showed bacterial spread, with corresponding graphs C,D) depicting the percentage inhibition of spread for E. coli ATCC 25 922 (A,C) and P. aeruginosa PAO1 (B,D) following treatment with different peptide concentrations (20 µm W‐W₁₃, 40 µm BTT1‐3A, 40 µm diF‐F₁₃, 100 µm E‐E₁₃ ^ASYM). After 30 minutes of incubation for nanonet formation, 2 µL of each mixture were seeded in 0.3% LB soft agar and the spreading of bacteria was investigated after 14 hrs. Experiments were performed in triplicates, with data expressed as mean ± SE. Statistical significance compared to the untreated control was evaluated using Dunnett's multiple comparisons (ns: p > 0.05, *: p ≤ 0.05, **: p ≤ 0.01, ****: p ≤ 0.0001).

Using an E. coli ΔflhD mutant with impaired motility, we demonstrated that without bacteria movement to escape from areas with high local drug concentrations or from loosely trapped nanonets, non‐motile bacteria were more susceptible to killing by polymyxin B or the loose BTT1‐3A nanonets (Figure S12, Supporting Information). W‐W₁₃ and diF‐F₁₃ effectively prevented the growth of non‐motile bacteria. Lastly, we confirmed that E‐E₁₃ ^ASYM nanonets without antibacterial activity could not prevent the bacterial outgrowth within the nanonets, posing a risk of potential spread. Since bacterial motility can significantly reduce the effectiveness of potent antibiotics,^{[ 3} , ^{4 ]} focusing solely on developing more potent antibiotics may not be the most effective approach. Instead, creating nanonets that tightly encage motile bacteria and restrict their movement could offer a promising strategy to prevent bacterial spread and infection.

3. Conclusion

In summary, we explored the design principles of peptide nanonets that enabled control over nanofiber interweaving on bacterial membranes to form distinct morphologies. This was achieved through rational modification of β‐hairpin side strands by leveraging the unique chemical properties of amino acid side chains. By carefully tuning the termini and aromaticity of the hydrophobic residue, W‐W₁₃ formed densely interwoven nanofibers on bacterial surfaces, creating tight nanonets that effectively reduced bacterial spread. In contrast, asymmetrical glutamic acid substitution produced E‐E₁₃ ^ASYM, which formed extensive self‐interwoven nanonets with minimal bacterial coverage and no antibacterial activity. Thus, controlling peptide self‐assembly is crucial for developing nanonets that tightly encage bacteria and prevent infection. Although only canonical amino acids were used in this study, this strategy can be extended to non‐canonical amino acid libraries to further enhance potency, proteolytic stability, and cytocompatibility, potentially addressing specific clinical needs.

4. Experimental Section

Materials

Cation‐adjusted Mueller Hinton broth (MHB), Luria broth base, Miller's modified (LB), LB Broth (Luria low salt), LPS from E. coli O111:B4, LTA from S. aureus, Vancomycin hydrochloride, Polymyxin B sulfate, Kanamycin sulfate, Penicillin‐Streptomycin solution, Propidium Iodide (PI), Trypsin‐EDTA solution, Dimethyl sulfoxide (DMSO) and Glutaraldehyde (grade I, for use as electron microscope fixative) were obtained from Sigma (St. Louis, MO). Osmium tetroxide (OsO₄) was purchased from Ted Pella (Redding, CA). K114 dye was obtained from Santa Cruz (Dallas, TX). 10X Phosphate Buffered Saline (PBS) was obtained from Vivantis Technologies Sdn Bhd (Shah Alam, Malaysia). LB Agar (Lennox L Agar) and BODIPY TR Cadaverine were purchased from ThermoFisher Scientific (Waltham, MA). HyClone Standard Fetal Bovine Serum (FBS) was obtained from Cytiva (Marlborough, MA). CellTiter‐Glo Luminescent Cell Viability Assay kit was purchased from Promega (Madison, WI). Triton X‐100 was obtained from Bio‐Rad Laboratories (Hercules, CA).

The peptides used in this study were synthesized by GL Biochem (Shanghai, China) and were confirmed to have a purity of at least 95%, as determined by the manufacturer using reversed‐phase high‐performance liquid chromatography (RP‐HPLC). Peptide characterization data were summarized in Table S3 (Supporting Information), and the corresponding analytical HPLC chromatograms and ESI(+) mass spectra were provided in Figures S13–S24 (Supporting Information). For all peptides, the differences between the observed and theoretical molecular weights for each charged state deviated by less than 1 Da, confirming the accuracy and reliability of the measurements.

Bacteria Strains

E. coli ATCC 25 922, S. aureus ATCC 29 737, and P. aeruginosa PAO1 were obtained from American Type Culture Collection (Manassas, VA). E. coli K12 BW25113 was obtained from Horizon Discovery (Waterbeach, UK).

Mammalian Cell Line

HaCaT keratinocytes were kind gift by Professor Paul Ho Chi Lui (NUS, Singapore). The cells were cultured in high‐glucose DMEM supplemented with 10% (v/v) FBS and 1% (v/v) penicillin‐streptomycin.

Prediction of Peptide Secondary Structures

The amino acid sequences of the designed peptide libraries were submitted to D‐I‐TASSER (https://zhanggroup.org/D‐I‐TASSER/) to predict their secondary structures.^{[ 30 ]} The predicted structures were then visualized using PyMol tool.

Circular Dichroism (CD) Spectroscopy

CD spectroscopy was used to evaluate the secondary structures of peptides under various treatments, following the method previously described.^{[ 32 ]} In brief, CD spectra were recorded for each peptide at a concentration of 50 µm, either in 1× PBS buffer (pH 7.4) alone or in the presence of 0.2 mg mL⁻¹ of LPS or LTA. The measurements were conducted using a quartz cell with a 1 mm path length over a wavelength range of 190–260 nm, with a scanning speed of 60 nm min⁻¹ on a Jasco CD Spectrometer (JASCO, Easton, MD). Each sample was measured three times, and the spectra were averaged. Mean residue ellipticity (θ_M ) was calculated using the following equation:

$${\theta }_M=\frac{{\theta }_{obs}}{10}.\frac{M_{RW}}{c.l}$$ (1)

θ_M represents the mean residue ellipticity (deg cm² dmol⁻¹), θ_obs was the observed ellipticity corrected for the blank at a given wavelength (mdeg), M_RW was the residue molecular weight (MW divided by the number of amino acids), c was the peptide concentration (mg mL⁻¹), and l was the path length (cm).

K114 Fluorescence Assay

The fibrillation of peptides was quantified by assessing the binding of K114 fluorescence dye. A 10 mM K114 stock solution was prepared by dissolving the dye in DMSO. This stock was then diluted to 250 µm with sterile deionized water, then immediately mixed and incubated with 20 µm of peptides in the presence of 5 µg mL⁻¹ LPS, LTA, or 1× PBS (pH 7.4) alone, at 37 °C with shaking at 120 rpm. The fluorescence intensity of K114 (λ_EX = 360 nm, λ_EM = 550 nm) was measured at 0.5, 1.0, and 2.0 hours using a TECAN Infinite M200 plate reader (Männedorf, Switzerland). The percentage increase in the K114 signal was calculated using the following equation:

$$PercentageincreaseinK114signal=\frac{\left(FluorescenceIntensit{y}_{\left(WithLPSorLTA\right)}-FluorescenceIntensit{y}_{\left(WithoutLPSorLTA\right)}\right)}{FluorescenceIntensit{y}_{\left(WithoutLPSorLTA\right)}}\times 100\%$$ (2)

BTT1‐3A served as positive control whereas BTT4 that did not fibrillate serves as negative control. Peptides that have higher fluorescence signal than BTT4 were identified as fibrillating peptides and selected for subsequent experiments.

Confocal Microscopy

To verify the spontaneous self‐assembly of diF‐F₁₅, a solution of 64 µm diF‐F₁₅ and 20 µm K114 dye was prepared in 1× PBS buffer (pH 7.4) and incubated at 37 °C with shaking at 120 rpm for 30 minutes. The sample was then placed on a coverslip, and z‐stack confocal images were captured at 5 µm intervals using a BC43 Andor confocal microscope to visualize the bound K114 dye with an excitation wavelength of 360 nm and an emission wavelength of 550 nm. The images were processed and visualized using Imaris software, with scale bars added using ImageJ software.

Field‐Emission SEM

The preparation of SEM samples was conducted as described in this previous work,^{[ 20 ]} with some minor modifications. Bacterial cultures in the mid‐log phase were diluted in MHB to an initial OD₆₀₀ of 0.07. Peptides were then added at a concentration of 40 µm, except 100 µm for E‐E₁₃ ^ASYM to the diluted bacteria culture at a final volume of 500 µL in 1.5 mL Eppendorf tubes. The mixtures were incubated at 37 °C with shaking at 120 rpm for 2 hours to induce nanonet formation. After incubation, the samples were centrifuged at 6000 g for 6 minutes, and the pellets were washed gently with 1× PBS (pH 7.4), a process repeated three times. The samples were then fixed overnight at 4 °C in 2.5% glutaraldehyde (in water). Following fixation, the samples were washed twice with 1× PBS and resuspended in 500 µL of 1× PBS. The samples were then added to pre‐coated poly‐L‐lysine coverslips for 30 minutes, and unbound bacteria were removed by washing with 1× PBS. The samples were post‐fixed with 1% OsO₄ for 30 minutes and washed twice with distilled water to remove any residual OsO₄. The samples were then dehydrated through a series of ethanol washes (50%, 75%, 95%, and three times with 100%). Finally, the samples were sputter‐coated with gold using a Leica EM ACE200 (Wetzlar, Germany) and imaged with a JEOL JSM‐6701F scanning electron microscope (Tokyo, Japan). At least three independent fields were imaged for each sample to ensure representativeness. The final images were processed in ImageJ (National Institutes of Health, Bethesda, MD) to include scale bars.

Minimum Inhibitory Concentration (MIC) assay

The broth microdilution method was used to determine the MICs of the peptides. Stock solutions of the peptides and the positive control antibiotics (vancomycin and polymyxin B) were first diluted to twice the highest tested concentration in a 96‐well plate using MHB. The peptides were then serially diluted two‐fold to achieve the desired concentration range, with a final volume of 50 µL per well. Bacterial cultures in the mid‐log phase were diluted to an OD₆₀₀ of 0.07, then further diluted 100‐fold in MHB to obtain an initial inoculum of 10⁶ colony‐forming units (CFU) mL⁻¹. A 50 µL aliquot of this bacterial suspension was added to each well containing the peptides, followed by incubation at 37 °C with shaking at 120 rpm for 18 hours. After incubation, the OD₆₀₀ was measured using a Tecan Infinite M200 microplate reader (Männedorf, Switzerland). The MIC value was defined as the lowest peptide concentration that resulted in a more than 90% reduction in OD₆₀₀ compared to the drug‐free control.

BODIPY‐Cadaverine (BC) Displacement Assay

The experimental protocol was adapted from a previously described method.^{[ 33 ]} The relative binding affinities of the peptides to bacterial LPS or LTA were assessed by evaluating their capacity to displace BC from LPS‐BC or LTA‐BC complexes. Displacement of BC, which binds strongly to LPS and LTA, results in dequenching and fluorescence of the free BC. Initially, 50 µg mL⁻¹ of LPS or LTA was incubated with 5 µm BC in 1× PBS buffer (pH 7.4) for 30 minutes at room temperature, protected from light. The peptides and polymyxin B control were prepared by diluting 1 mM stock solutions to twice the highest tested concentration in a Costar 96‐well black polystyrene plate using 1x PBS buffer. Serial two‐fold dilutions were then performed to achieve the desired concentration range, with a final volume of 50 µL per well. Subsequently, 50 µL of the LPS/LTA‐BC mixtures were added to the wells containing the peptides and incubated at 37 °C with shaking at 120 rpm for 30 minutes. After incubation, the fluorescence intensity of the BC samples (λ_EX = 580 nm, λ_EM = 620 nm) was measured using a TECAN Infinite M200 plate reader (Männedorf, Switzerland). The BC Occupancy Factor was calculated using the following equation:

$$BCOccupancyFactor=\frac{(F_{max}-F_{sample})}{\left(F_{max}-F_{min}\right)}$$ (3)

where F_max represents the highest fluorescence intensity signal of among the samples, F_min the lowest and F_sample the fluorescence intensity signal of the test sample.

The percentage of BC displaced was calculated using equation below:

$$PercentageofBCdisplaced\left(\%\right)=\left(1-BCOccupancyfactor\right)\times 100\%$$ (4)

Agglutination Assay

This experiment followed previously established, optimized conditions to facilitate comparison with the parent peptide BTT1‐3A.^{[ 20 ]} Briefly, the bacteria were initially cultured in MHB overnight to achieve a sufficient density. The cultures were then centrifuged at 3000 rpm for 7 minutes, after which the supernatant was discarded, and the bacterial pellet was resuspended in 1× PBS buffer (pH 7.4). The bacterial suspension was further diluted with 1× PBS to obtain a starting OD₆₀₀ of 0.23, measured using a disposable plastic cuvette (12 × 12 × 4.5 mm) with 1 mL of the bacterial culture. A 1 mM stock solution of the peptide was then added to the bacterial suspension to achieve a final peptide concentration of 40 µm in a total volume of 1 mL. After thorough mixing, 1 mL of the sample was transferred to a disposable plastic cuvette, and the OD₆₀₀ was measured immediately at time T = 0 h. The OD₆₀₀ was subsequently recorded at hourly intervals for up to 8 hours at room temperature using a Tecan Infinite M200 microplate reader (Männedorf, Switzerland).

Cell Viability Assay

HaCaT keratinocytes were seeded at a density of 10000 cells per well in a 96‐well plate and allowed to adhere for 24 hours at 37 °C with 5% CO₂. After this period, the culture medium was replaced with fresh serum‐free media containing peptides at desired concentrations of up to 1 mM. For controls, 1% Triton X was used as a positive control, while serum‐free medium served as a negative control. Following 24 hours of incubation, the cells were washed, and 110 µL of a 10× diluted CellTiter‐Glo reagent (Promega) in serum‐free media was added to each well, followed by a 3.5‐hour incubation at 37 °C with 5% CO₂. Absorbance was then measured at 490 nm using a Tecan Infinite M200 microplate reader (Männedorf, Switzerland). The absorbance reading from the 1% Triton X control was subtracted from each sample as a blank. The percentage of cell viability was calculated using the following equation:

$$Percentagecellviability\left(\%\right)=\frac{Absorbanc{e}_{Peptide-treatedcells}}{Absorbanc{e}_{Untreatedcells}}\times 100\%$$ (5)

Cytotoxicity curves were plotted, and CC₅₀ values were determined using non‐linear regression analysis with GraphPad Prism 9.5 software.

HaCaT Keratinocyte Membrane Damage Assay

HaCaT keratinocytes were seeded at a density of 10000 cells per well in a 96‐well plate and allowed to adhere for 24 hours at 37 °C with 5% CO₂, reaching 70–80% confluency. The culture medium was then replaced with fresh serum‐free media containing peptides at 25 µm, and the cells were incubated for 1 hour under the same conditions. It was previously optimized in this study that 25 µm BTT1‐3A as the lowest concentration causing significant membrane disruption, hence this concentration was selected for testing all peptides in this work. Following incubation, the cells were washed once with 1× PBS buffer (pH 7.4) and stained with a 50 µg mL⁻¹ PI solution in 1× PBS at 37 °C for 15 minutes. After staining, the cells were washed twice with 1× PBS, and serum‐free media were added to each well. The EVOS M7000 Imaging System (Life Technologies Corporation, Bothell, WA) was used to observe cell morphology and PI‐stained nuclei. Final images were processed in ImageJ (National Institutes of Health, Bethesda, MD) to include scale bars.

Soft Agar Motility Assay

All bacterial strains were cultured in LB medium until reaching mid‐log phase, except for non‐motile E. coli ΔflhD mutant, which was cultured in low‐salt LB supplemented with 25 µg mL⁻¹ kanamycin. The bacterial cultures were then diluted to 10⁷ CFU mL⁻¹ in LB and incubated with peptides at the desired concentrations in a final volume of 60 µL in 1.5 mL Eppendorf tubes. The samples were thoroughly mixed by pipetting and incubated at 37 °C with shaking at 120 rpm for 30 minutes. Subsequently, 2 µL of each sample was seeded into 0.3% w/v LB agar plates and incubated at room temperature for 14 hours. After incubation, images of the agar plates were captured. Each sample was performed in independent triplicate. The diameter of the bacterial spreading zones on the plates was measured using ImageJ (National Institutes of Health, Bethesda, MD). The percentage inhibition of bacterial spreading was calculated using the following equation:

$$\begin{array}{ccc}& & PercentageInhibition\left(\%\right)\hfill \\ & & =\frac{Diamete{r}_{\left(Untreatedbacteria\right)}-Diamete{r}_{\left(Peptide-treatedbacteria\right)}}{Diamete{r}_{\left(Untreatedbacteria\right)}}\times 100\%\hfill \end{array}$$ (6)

Statistical Analysis

Differences between groups were analyzed using one‐way or two‐way ANOVA, followed by Dunnett's multiple comparisons test to compare each experimental group to the control group. (ns: p > 0.05, *: p ≤ 0.05, **: p ≤ 0.01, ***: p ≤ 0.001, ****: p ≤ 0.0001). GraphPad Prism 9.5 was used for statistical tests, graph drawing, and curve fitting.

Conflict of Interest

The authors declare no conflict of interest.

Supporting information

Supporting Information

Chen W. M., Murali D. M., Tram N. D. T., Yu P., Wang D. T., and Ee P. L. R., “Controlling Nanonet Morphology via Residue‐Specific Modulation of β‐Hairpin Peptide for Enhanced Bacterial Trapping.” Small 21, no. 35 (2025): 21, 2505823. 10.1002/smll.202505823

Data Availability Statement

The data that supports the findings of this study are available in the supplementary material of this article.