A dual-targeting lipid nanoparticle delivery system for precision phage therapy against multidrug-resistant Gram-negative bacteria

Abstract

The escalating crisis of antimicrobial resistance, particularly among Gram-negative bacteria, demands innovative therapeutic strategies. Here, we present a dual-targeting lipid nanoparticle-chimeric phage (AbLCP) system engineered to combat multidrug-resistant (MDR) Gram-negative bacteria. Our design integrates a genetically modified M13 phage displaying anti–lipopolysaccharide (LPS) antibodies with polymyxin B conjugation, which is further encapsulated within anti-LPS antibody–functionalized lipid nanoparticles. This system overcomes the fundamental limitations of conventional phage therapy, enabling efficient cellular internalization, endosomal escape, and potent intracellular bactericidal activity. In murine models of Pseudomonas aeruginosa infection, AbLCP achieved superior bacterial clearance and up to 10³-fold reduction in bacterial burden at a 10³-fold lower effective dose than free polymyxin B, with 67-fold selective accumulation at infection sites, and attenuated systemic inflammation and preservation of gut microbiota. Mechanistically, AbLCP neutralized LPS to suppress inflammation and induced a lethal bacterial stress response. This work establishes a transformative nanophage platform that merges synthetic biology with nanomedicine, offering a promising strategy against MDR bacterial infections.

INTRODUCTION

Antimicrobial resistance has become one of the most critical challenges to modern medicine. The World Health Organization estimates that multidrug-resistant (MDR) pathogens cause more than 700,000 deaths each year, with projections reaching 10 million by 2050 if effective interventions are not implemented (1–3). Among these pathogens, Gram-negative bacteria such as Pseudomonas aeruginosa (4–7), Klebsiella pneumoniae (8–11), and Acinetobacter baumannii represent the highest priority group because of their notable ability to resist nearly all classes of antibiotics (12–19). Their resistance arises from a combination of intrinsic and acquired mechanisms, including an impermeable outer membrane enriched in lipopolysaccharides, constitutively active efflux pumps, and the formation of biofilms (20–24). These biological barriers make Gram-negative bacterial infections particularly recalcitrant in hospital and community settings (17, 25–27). The crisis is further exacerbated by the overuse of broad-spectrum antibiotics and the nearly stagnated discovery pipeline for new small-molecule antibiotics (28–30).

Bacteriophage (or phage) therapy has reemerged as a promising alternative to conventional antibiotics (31–33). Phages are natural bacterial viruses that selectively infect and lyse their host bacteria while sparing commensal microbes (34–36). However, multiple translational barriers impede the reliable use of phages in clinical settings. Free phages are rapidly inactivated by harsh conditions, while their size and charge impede biofilm and tissue penetration, and they generally fail to access critical intracellular bacterial reservoirs (37–39). Furthermore, free phages are quickly cleared from systemic circulation by the reticuloendothelial system and can elicit neutralizing immune responses upon repeated administration (40–42). Highly lytic phages, while potent, may also exacerbate inflammation through uncontrolled bacterial lysis and massive release of endotoxin (43–46). To address these issues, recent advances in synthetic biology and nanotechnology have opened unique opportunities to reengineer phage systems for enhanced stability, targeting, and therapeutic precision (47–51). The nonlytic filamentous M13 phage, in particular, has emerged as a favorable platform for genetic and chemical modification (52, 53). However, the clinical translation of M13-based phage therapy remains challenging. Circulating M13 particles are subject to rapid clearance and degradation (40), and their size (∼900 nm in length) further restricts diffusion through mucus layers, extracellular matrices, and biofilm architectures (54–56). Overcoming these physicochemical barriers is thus a prerequisite for realizing the full potential of M13-derived therapeutic platforms. Lipid nanoparticles (LNPs) have revolutionized the field of nanomedicine by enabling efficient delivery of biomacromolecules through their protective lipid bilayer and tunable physicochemical properties (57–60). However, existing studies have primarily used passive encapsulation approaches to enhance phage stability, lacking active targeting mechanisms (61). Moreover, conventional LNPs have not been optimized to penetrate bacterial biofilms or access intracellular infection niches, two major reservoirs responsible for bacterial persistence and relapse.

Facing these challenges, herein, we reported a double-targeting LNP chimeric phage delivery system (AbLCP). The aim of this study is to develop a precise phage therapy system with enhanced translational potential, specifically enhanced biofilm clearance and intracellular bactericidal ability, as well as resolved inflammatory response through synthetic biology and nanotechnology. To avoid undesirably uncontrolled release of bacterial endotoxins, we used a well-characterized and sequence-programmable M13 phage. The wild-type M13 phage is a temperate phage, infecting F⁺ Escherichia coli with a nonlethal life cycle and very low bactericidal efficiency (62). To extend the spectrum and enhance the antibacterial capability, the chimeric M13 phage (CM13) is engineered to display anti-LPS antibodies on its g3p coat protein for specific recognition of Gram-negative bacteria (63) and to conjugate the antibiotic polymyxin B (PMB) onto its g8p major coat protein (64). The filamentous phage is confined into the LNPs with anti-LPS antibodies decorated on the surface. This dual-targeting architecture allows AbLCP to engage with bacterial surfaces through multivalent interactions, traverse dense biofilm matrices, and be internalized by infected epithelial or immune cells, enabling simultaneous elimination of both extracellular and intracellular bacterial reservoirs. Notably, it can also protect the phages from harsh in vivo conditions and modulate the local microenvironment to mitigate LPS-induced inflammatory responses (65). AbLCP exhibits superior antibacterial performance in P. aeruginosa infection models, providing a blueprint for constructing next-generation therapeutic systems that integrate phage biology with nanomedicine.

RESULTS

Construction and characterization of AbLCP

The AbLCP particles were constructed by encapsulating the CM13 into LNPs through extrusion, followed by surface functionalization with anti-LPS antibodies (Fig. 1A). To extend the host spectrum, we genetically engineered the phage to replace its N-terminal domain of the receptor binding protein with anti-LPS antibodies. Then, the g8p capsid proteins (∼2700 copies) on the side wall were covalently conjugated with PMB through 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide (EDC)/N-hydroxysuccinimide (NHS) chemistry. Each g8p protein bears a minimum of three solvent-accessible carboxyl-containing residues (Glu², Asp⁴, and Asp⁵) at the N terminus. These groups can actively react with the amine groups of PMB under mild conditions. By elegantly tuning the concentrations of the reagents, the morphology of the obtained CM13 remains unaltered without cross-linking or aggregation, as observed in the transmission electron microscopy (TEM) and atomic force microscopy (AFM) images (Fig. 1B and fig. S3, A to C). After extrusion through a 200-nm polycarbonate membrane, the LNP-encapsulated chimeric phage (LCP) nanoparticles with controlled and uniformed size were obtained. AbLCP was fabricated by coupling the surface NHS groups with the amines of anti-LPS antibodies, verified by SDS–polyacrylamide gel electrophoresis (SDS-PAGE) and Western blot analysis (fig. S1).

Fig. 1. Construction and characterization of AbLCP.

(A) Schematic illustration of AbLCP construction. (B) AFM image of CM13. (C and D) AFM (C) and cryo-TEM (D) images of AbLCP. (E) SIM images of AbLCP indicating successful encapsulation of CM13 inside the lipid nanoparticle; scale bars, 100 nm in (D) and (E). (F) Size distribution of LCP and AbLCP measured by DLS. (G) Zeta potential of M13, CM13, LCP, and AbLCP measured by Zetasizer. (H) Zeta potentials and hydrodynamic diameters of AbLCP over 28 days of storage in 4°C. Data are presented as the means ± SEM (n = 3). Statistical analyses were performed using a one-way ANOVA followed by Tukey’s post hoc test. ***P < 0.001.

AFM, cryo–transmission electron microscopy (cryo-TEM), structured illumination microscopy (SIM), and dynamic light scattering (DLS) analysis revealed that encapsulated CM13 phages adopted a curled conformation within LNPs, reducing their effective size from ∼900 nm (free phage) to a hydrodynamic diameter of ∼200 nm (Fig. 2, C to F, and figs. S2 and S3, A to D). This reversible physical deformation does not cause permanent structural damage or functional loss. Benefiting from the high flexibility of filamentous phage (66), CM13 can fully restore its antibacterial activity once released from LNPs (figs. S10 and S12). The PMB content in CM13 [10⁹ plaque-forming units (PFU)/ml] was ∼0.05 μg/ml, as determined by liquid chromatography–mass spectrometry (LC-MS). AbLCP exhibited a positive zeta potential of 37.5 mV (Fig. 1G), indicating a positively charged surface that may facilitate cellular internalization. AbLCP demonstrated robust stability after 28-day storage at 4°C, with no pronounced change in size, zeta potential, and antibacterial ability (Fig. 1H and fig. S4A). It maintained structural integrity across a broad pH range (2.5 to 11.5; fig. S4, B, D, and E), suggesting suitability for oral delivery. Notably, exposure to phospholipase, a key virulence factor in bacterial biofilms, triggered particle deformation and subsequent degradation, providing a mechanism for microenvironment-responsive cargo release (fig. S4, C and F).

Fig. 2. Biofilm penetration capability of AbLCP.

(A to C) Binding assay of wild-type M13 (A), CM13 (B), and AbLCP (C) to P. aeruginosa measured by surface plasmon resonance (SPR). (D) Representative three-dimensional reconstructions from confocal image stacks of green fluorescent protein (GFP)–expressing P. aeruginosa biofilms, together with their merged channels, following 0.5 hours of treatment with CM13, LCP, or AbLCP. Red signals indicate PE-labeled phage CM13, and green signals indicate the P. aeruginosa biofilms. Scale bar, 50 μm. (E) z-Axis cross-sectional views showing the penetration of CM13, LCP, and AbLCP. Scale bar, 100 μm. (F) Fluorescence quantification of the penetration depth of CM13, LCP, and AbLCP. a.u., arbitrary units. Data are presented as the means ± SEM (n ≥ 3).

Enhanced biofilm penetration capability of AbLCP

The dual-targeting architecture allows AbLCP to capture a wider variety of Gram-negative bacteria by attaching the LPS molecules on the cell outer membrane. Binding affinity assays confirmed the high affinity of AbLCP and CM13 to LPS [AbLCP, dissociation constant (K_d) = 1.26 × 10⁻¹⁵ M; CM13, K_d = 1.04 × 10⁻¹² M; fig. S5A] and typical Gram-negative bacteria (Fig. 2, A to C, and fig. S5, B to D). Cryo-TEM and confocal microscopy verified that the anti-LPS antibody successfully endowed CM13 and AbLCP with specific targeting capabilities, facilitating robust colocalization with both planktonic and biofilm-embedded pathogens, including P. aeruginosa, K. pneumoniae, A. baumannii, and Enterobacter cloacae (Fig. 2D and figs. S3 and S6 to S8). In addition, typical drug-resistant Gram-negative bacteria were treated with phycoerythrin (PE)–labeled CM13 and AbLCP. Their capturing abilities to these pathogens were verified by flow cytometry (fig. S9). These tests confirmed that the host spectrum of the phage system is successfully extended through the interaction of the antibody and LPS. It could benefit the penetration behavior in the dense biofilm of Gram-negative bacteria in which LPS is widely distributed.

Time-kill assays of Gram-negative pathogens demonstrated that AbLCP and CM13 rapidly reduced viable bacterial counts by several orders of magnitude within 12 hours, substantially more efficient than using PMB (fig. S10). Then, P. aeruginosa biofilms were treated with PE-labeled AbLCP, LCP, or naked CM13 to assess the penetration capability. AbLCP exhibited uniform fluorescence throughout the full biofilm depth (∼150 μm), whereas LCP penetrated 40 to 90 μm and naked CM13 remained largely superficial (0 to 50 μm) (Fig. 2, D to F). The transport efficiency was evaluated by a transwell biofilm model. Recovered phage titers followed AbLCP > LCP > CM13 > wild-type M13, with AbLCP yielding a 4.2-fold increase over CM13 (fig. S11). The enhanced biofilm penetration and more efficient eradication of bacterial pathogens embedded in the biofilm (fig. S12) result from combined advantages of the AbLCP system. The smaller spherical structure allows more efficient diffusion through biofilm interstitial gaps (67). The surface anti-LPS antibodies achieve specific targeting and prolonged retention inside biofilms, converting passive diffusion into active targeting that extends particle retention and drives deeper infiltration (68, 69). Furthermore, the lipid bilayer shields encapsulated CM13 from enzymatic degradation, and the positive surface charge facilitates attachment and penetration via electrostatic attraction to negatively charged biofilms. These complementary features jointly support the outstanding performance of AbLCP.

Cellular uptake and intracellular trafficking of AbLCP

We next explored whether LNP encapsulation could facilitate the entry of CM13 into eukaryotic cells by incubating human embryonic kidney (HEK) 293 cells with PE-labeled AbLCP or naked CM13. Confocal imaging revealed that negligible fluorescence was detected in cells treated with naked CM13 even after 4 hours (Fig. 3A). In contrast, weak but distinct red fluorescence was observed on the surface of HEK293 cells after only 1 hour of incubation with AbLCP, likely due to the enhanced interactions with the cell membrane via hydrophobic and electrostatic forces. The fluorescence signals migrated from the cell surface to the cytoplasm at 2 hours and were distributed throughout the cell interior at 4 hours, demonstrating efficient and complete internalization of AbLCP into HEK293 cells.

Fig. 3. Entry into HEK293 cells and lysosomal escape of AbLCP.

(A) Representative confocal images of HEK293 cells (5 × 10⁴ cells per well) after incubation with naked CM13 (1 × 10⁹ PFU/ml) or AbLCP (1 × 10⁹ PFU/ml equivalent to CM13) for 1, 2, and 4 hours (h), respectively. Blue indicates Hoechst-stained cell nuclei, and red indicates PE-labeled phages. Scale bar, 10 μm. (B) Corresponding flow cytometry histograms of HEK293 cells. Phages were labeled with PE, and untreated cells were used as the negative control. (C) Typical confocal images of HEK293 (5 × 10⁴ cells per well) following incubation with CM13 (1× 10⁹ PFU per ml) or AbLCP (1 × 10⁹ PFU/ml equivalent) for 1 and 2 hours, respectively. Phages were labeled with PE, and cell nuclei and endosomes were stained with Hoechst and LysoTracker Green, respectively. Scale bar, 5 μm.

Consistent with confocal imaging results, flow cytometry quantification confirmed minimal uptake of naked CM13, with mean fluorescence intensity (MFI) similar to that of untreated cells (Fig. 3B and fig. S13). In contrast, AbLCP-treated cells showed a time-dependent increase in MFI. Their MFI was remarkably higher than that of naked CM13–treated cells at 1 hour, and this difference became more pronounced at 2 and 4 hours. Notably, the MFI of AbLCP-treated cells at 4 hours was ∼68-fold higher than that of naked CM13–treated cells (fig. S13). Furthermore, intracellular trafficking studies revealed that naked CM13 phages showed no obvious colocalization with lysosomal markers at any time point (Fig. 3C), consistent with their poor cellular internalization (Fig. 3A). On the contrary, AbLCP initially colocalized with lysosomal markers at 1 hour, confirming endocytic uptake. By 2 hours, colocalization decreased and phage signals dispersed throughout the cytosol, indicating efficient endosomal/lysosomal escape. This shift indicates that LNP encapsulation not only promotes cellular uptake of CM13 but also facilitates efficient endosomal/lysosomal escape. This stark contrast in fluorescence signals between AbLCP and CM13 clearly demonstrated that LNP encapsulation markedly enhances the cellular uptake efficiency of CM13 in HEK293 cells. The naked CM13 phage is negatively charged and fails to enter host cells because of electrostatic repulsion with negatively charged cell membranes. After LNP encapsulation, AbLCP acquires a positive surface charge (Fig. 1G). The resulting electrostatic attraction between positively charged AbLCP and host cell membranes, combined with membrane fusion mediated by the lipid shell, jointly enables efficient cellular internalization (70, 71). AbLCP interacts with the endosomal membrane via electrostatic and hydrophobic forces, destabilizing the membrane and deforming the particle. Endosomal phospholipases and lipid-interacting proteins further disrupt the LNP bilayer. These combined stresses compromise AbLCP stability, eventually releasing the encapsulated CM13. This efficient endosomal escape underscores the platform’s potential for targeted intracellular antibacterial therapy.

Enhanced abilities of AbLCP in intracellular bacterial inhibition

To evaluate intracellular antibacterial activity, HEK293 cells were infected with mCherry-expressing P. aeruginosa (fig. S14) and treated with carbenicillin to remove extracellular bacteria before adding PMB, CM13, LCP, or AbLCP. Confocal imaging revealed strong intracellular fluorescence in phosphate-buffered saline (PBS)–, PMB-, and CM13-treated cells, indicating persistent infection and negligible antibacterial effect of PMB and CM13 against intracellular bacteria (Fig. 4, A to C). In contrast, LCP- and AbLCP-treated cells showed minimal bacterial signal and maintained healthy morphology, suggesting effective bacterial clearance and restored cell viability (Fig. 4, D and E). Flow cytometry confirmed that naked CM13 did not reduce bacterial load, while LCP and AbLCP decreased MFI by 60.6 and 65.3%, respectively (P < 0.0001; Fig. 4, F and G), suggesting efficient bacterial clearance. Beyond HEK293 epithelial cells, THP-1 macrophages were selected as an additional cell model to assess intracellular antibacterial performance (fig. S15). As professional phagocytes, macrophages are key immune cells and major intracellular reservoirs for numerous pathogenic bacteria (72–75). These cells harbor a complex, hostile intracellular environment that challenges the activity of antibacterial agents. This model closely recapitulates the physiological state of intracellular bacterial parasitism in vivo. Effective bacterial clearance in THP-1 macrophages confirms that AbLCP can eradicate pathogens residing within immune cell reservoirs, inhibiting bacterial persistence and dissemination. Together, data from two distinct cell types fully verify the broad-spectrum intracellular antibacterial capacity and practical application potential of AbLCP against intracellular bacterial infections.

Fig. 4. Intracellular bacterial inhibition of AbLCP.

(A to E) Representative confocal images of mCherry-expressing P. aeruginosa–infected HEK293 cells (5 × 10⁴ cells per well) after 2 hours of treatment with (A) PBS, (B) PMB (2 μg/ml), (C) CM13 [1 × 10⁹ PFU/ml equivalent to PMB (0.05 μg/ml)], (D) LCP (1 × 10⁹ PFU/ml equivalent), or (E) AbLCP (1 × 10⁹ PFU/ml equivalent), respectively. BF, bright field. Red signal indicates mCherry-expressing P. aeruginosa. Scale bar, 5 μm. (F and G) Flow cytometry histograms (F) and MFI (G) of HEK293 cells (5 × 10⁴ cells per well) following incubation with PBS, PMB, CM13 (1 × 10⁹ PFU/ml), or AbLCP (1 × 10⁹ PFU/ml equivalent). (H and I) Photo (H) and counts (I) of intracellular mCherry-expressing P. aeruginosa after different treatments for 4 hours. Data are presented as the means ± SEM (n = 5). Statistical analyses were performed using a one-way ANOVA followed by Tukey’s post hoc test. ****P < 0.0001; ns, not significant, P > 0.05.

To quantify intracellular bacterial viability, the number of colony-forming units (CFUs) of mCherry-expressing P. aeruginosa recovered from lysed HEK293 cells was determined. The antibacterial efficacy followed a clear hierarchy: AbLCP > LCP > CM13 > PMB > PBS control. Notably, treatment with AbLCP resulted in a profound reduction of viable bacteria, with no detectable colonies observed on the plates, indicating complete eradication within the detection limit (Fig. 4, H and I). LCP and AbLCP have consistent physicochemical properties, except for surface anti-LPS modification. Anti-LPS antibodies do not facilitate cellular internalization of the LNP platform, as lipopolysaccharide is not present on eukaryotic cell surfaces (76–78). The stronger intracellular bacterial inhibition of AbLCP is mainly ascribed to its specific targeting capability toward bacteria, which elevates the local concentration of active agents around intracellular bacteria and improves the therapeutic effect. These results establish that LNP encapsulation overcomes the intrinsic cell-entry limitation of filamentous phages and enables potent intracellular bactericidal activity.

Superior in vivo therapeutic values of AbLCP

To assess the in vivo targeting specificity of AbLCP, a murine model of localized bacterial infection was established in the right hindlimb (fig. S16A). Following intravenous administration of Cy7-labeled AbLCP, whole-body fluorescence imaging revealed detectable signal at the infection site within 2 hours, with sustained accumulation over 24 hours and minimal signal in the contralateral uninfected tissue (fig. S16B). Dynamic quantification of fluorescence intensity of the infected versus uninfected limb tissues detailed the spatiotemporal accumulation profile (fig. S16C), demonstrating specific and prolonged retention of AbLCP at the bacterial focus. Ex vivo imaging at 24 hours confirmed predominant accumulation in the liver, spleen, and infected tissue (fig. S16, D and E). The fluorescence intensity at the infection site was ∼67-fold higher than in corresponding uninfected tissue, confirming the exceptional targeting precision of the formulation.

The in vivo antibacterial efficacy of AbLCP was first assessed using a P. aeruginosa–induced peritonitis model. AbLCP treatment substantially reduced both total and intracellular bacterial counts compared to PMB and CM13, while extracellular bacterial counts were markedly higher in the LCP group (Fig. 5, A to E). CM13 carries a negative surface charge and acts directly on extracellular P. aeruginosa in the peritoneal cavity, thus exerting strong extracellular bactericidal activity. Nevertheless, electrostatic repulsion between negatively charged CM13 and cell membranes blocks its cellular entry, leading to weak efficacy against intracellular bacteria. LCP has a positive surface charge derived from the lipid carrier. This property facilitates cellular internalization to deliver CM13 into cells for intracellular bacterial elimination, while the lipid shell shields CM13 from enzymatic degradation in vivo. Consequently, LCP shows inferior activity against extracellular bacteria but superior performance in clearing intracellular pathogens compared with free CM13. This observation is consistent with a pronounced reduction in neutrophil infiltration within the tissue of PMB-, CM13-, and AbLCP-treated mice compared to the LCP group and PBS control, as visualized by Ly6G immunofluorescence staining (fig. S17). Extracellular bacteria exposed in peritoneal fluid and on the peritoneal surface induce prominent neutrophil infiltration in infected peritoneal tissue. In contrast, intracellular bacteria evade extracellular immune detection and fail to trigger strong neutrophil chemotaxis. Consistent with enhanced bacterial clearance, AbLCP-treated mice showed dramatic reductions in pro-inflammatory cytokines interleukin-6 (IL-6) and CXCL1/KC (fig. S18). Immunofluorescence analysis further confirmed substantially lower residual bacterial burden in major organs (liver, spleen, and kidney) of the AbLCP group compared to all other treatments (Fig. 5, F to H). Notably, the equivalent dose of free PMB (70 μg per dose) is ∼500-fold higher than that of AbLCP (equivalent to 0.134 μg of PMB per dose), indicating considerably superior in vivo therapeutic values of AbLCP.

Fig. 5. In vivo antibacterial efficiency of AbLCP in P. aeruginosa–induced peritonitis mice.

(A) Schematics of the P. aeruginosa peritonitis model. (B to E) Representative photographs (B) and CFUs in the total (C), extracellular (D), and intracellular (E) fractions were measured 24 hours after the different treatments (PMB, 70 μg per dose; CM13, equivalent to 0.67 μg of PMB per dose; LCP, equivalent to 0.134 μg of PMB per dose; AbLCP, equivalent to 0.134 μg of PMB per dose). Data are presented as the means ± SEM (n = 5). Statistical analyses were performed using a one-way ANOVA followed by Tukey’s post hoc test. **P < 0.01; ****P < 0.0001; ns, not significant, P > 0.05. (F to H) Representative immunofluorescence images of (F) liver, (G) spleen, and (H) kidney sections from mice treated with PBS, PMB, CM13, LCP, or AbLCP after peritonitis therapy. Tissue sections were stained for F-actin (phalloidin, gray/white), P. aeruginosa (primary antibody ab68538 followed by Cy3-conjugated secondary antibody, red), and nuclei [DAPI (4′,6-diamidino-2-phenylindole), blue]. Scale bars, 10 μm.

Then, the therapeutic effect of AbLCP in a more complicated in vivo environment was further investigated in a murine model of intestinal P. aeruginosa infection (Fig. 6A). The harsh environment of the gastrointestinal tract, characterized by low pH in the stomach, digestive enzymes, bile salts, and the dense mucus layer, poses tremendous challenges to the stability and bioavailability of both antibiotics and phages. Naked CM13 is susceptible to rapid degradation by proteases and obstruction by mucus because of its bulky filamentous structure, thus impairing its capacity to access target bacteria in the intestinal niche. The lipid bilayer of LNPs provides a protective shell that shields CM13 from the harsh conditions of the upper gastrointestinal tract, ensuring its structural integrity until it reaches the intestine. Ex vivo fluorescence imaging showed substantially greater AbLCP signal in intestinal tissues at 150 min post–oral dose compared to CM13 (Fig. 6B). Upon reaching the intestinal site of bacterial infection, the LNP shell is specifically ruptured by enzymes like bacterium-derived phospholipase C, triggering the site-specific release of the encapsulated CM13. This targeted delivery not only enhances the local concentration of CM13 but also prolongs their residence time in the intestinal lumen, maximizing their interaction with P. aeruginosa biofilms and planktonic cells embedded in the mucus layer.

Fig. 6. High therapeutic value of AbLCP in intestinal P. aeruginosa infection.

(A) Experimental design for assessing the in vivo efficacy of AbLCP in a mouse intestinal infection model. (B) Representative ex vivo IVIS images (left) and quantitative analysis of intestinal tissues from mice treated with PE-labeled AbLCP and naked CM13 (equivalent to 0.134 μg of PMB per dose). (C to F) Quantitative analysis of bacterial burden (plate counting) of ileum (C and E) and colon (D and F) after treatment with PBS, PMB, CM13, LCP, or AbLCP (PMB, 140 μg per dose; CM13, equivalent to 0.134 μg of PMB per dose; LCP, equivalent to 0.134 μg of PMB per dose; AbLCP, equivalent to 0.134 μg of PMB per dose). (G and H) Levels of IL-6 (G) and TNF-α (H) in the serum of mice treated with PBS, PMB, CM13, LCP, or AbLCP. Data are presented as the means ± SEM (n = 5). Statistical analyses were performed using a one-way ANOVA followed by Tukey’s post hoc test. **P < 0.01; ****P < 0.0001.

Oral administration of AbLCP (equivalent to 0.134 μg of PMB per dose) resulted in a substantial reduction in bacterial burden by ∼1000-fold versus free PMB (140 μg per dose; Fig. 6, C to F). Dose-response studies confirmed the superior efficacy of AbLCP across a range of concentrations (figs. S19 and S20). Mice receiving AbLCP also showed rapid recovery of intestinal barrier integrity (fig. S21) and a notable reduction in the inflammatory cytokines IL-6 and tumor necrosis factor–α (TNF-α) (Fig. 6, G and H), consistent with effective bacterial clearance. These findings collectively underscore that LNP encapsulation provides a decisive advantage not only in enhancing systemic stability and bioavailability but, crucially, in facilitating cellular entry and potent intracellular antibacterial action, which is inaccessible to conventional therapies.

AbLCP restored the diversity of the gut microbiota in bacterial enteritis

To assess the impact of AbLCP treatment on gut microbiota, we performed 16S ribosomal RNA sequencing on fecal samples from all groups. Sufficient sequencing depth was achieved to capture the majority of microbial diversity across all samples (Fig. 7A). Analysis of α diversity revealed that the Shannon index was substantially increased in AbLCP-treated mice compared to the infection model group, trending toward the level observed in the healthy controls. Notably, the diversity in the AbLCP group was also greatly higher than that in PMB-treated mice (Fig. 7B). It reflects that AbLCP can minimize collateral damage to the commensal gut ecosystem because of the precise targetability. Principal coordinates analysis (PCoA) demonstrated a distinct separation of the AbLCP-treated group from the disease model group, although it did not fully overlap with the healthy control cluster (Fig. 7C). This finding indicates that AbLCP therapy did not simply revert the microbiota to its naive preinfection state but actively remodeled the severely dysbiotic community into an alternative, distinct, and stable ecological structure. In contrast, the PMB-, CM13-, and LCP-treated groups clustered closely with the model group and remained distant from the healthy controls, indicating that these treatments failed to rectify the infection-induced dysbiosis and instead maintained a disease-like community structure (fig. S22A).

Fig. 7. AbLCP restored gut microbial diversity and reshaped the community structure in bacterial enteritis.

(A) Rarefaction curves indicating sufficient sequencing depth to capture microbial diversity across samples. (B) Bar graph of the Shannon index, showing α diversity (***P < 0.001; ****P < 0.0001). (C) PCoA plot based on Bray-Curtis dissimilarity, illustrating β diversity and separation among the model, control, and AbLCP groups. (D) Venn diagram depicting the number of shared and unique ASVs among the model, control, and AbLCP groups. (E) Left: UPGMA cladogram showing the hierarchical clustering of samples based on community similarity. Right: Stacked bar plots of relative taxonomic abundance at the phylum level for the model, control, and AbLCP groups. (F and G) Stacked bar plots of relative taxonomic abundance for each group at the (F) phylum and (G) order levels. (H) Heatmap of species abundance (at the order level) across all samples, clustered to visualize patterns in community composition.

Venn diagram analysis of shared microbial features [amplicon sequence variants (ASVs)] among groups provided further insight. While 63 ASVs (23.95%) were common to all three groups, the AbLCP-treated group shared substantially more ASVs with healthy controls (105, 39.9%) than with the model group (76, 28.9%) or than the model group shared with controls (96, 36.5%) (Fig. 7D). This pattern confirms that the AbLCP group displayed a microbial composition most closely resembling that of healthy controls. Notably, the PMB-, CM13-, and LCP-treated groups shared fewer ASVs with healthy controls than the AbLCP group, confirming the superior ability of AbLCP to restore a healthy-like microbial composition (fig. S22C). Consistent with this, unweighted pair group method with arithmetic mean (UPGMA) clustering showed high similarity between the AbLCP and healthy control groups at the phylum level (Fig. 7E). For the other treatments, similarity to healthy controls ranked as CM13 > LCP > PMB, but none reached the level of healthy mice (fig. S22B). Furthermore, relative abundance bar plots at the phylum and order levels (Fig. 7, F and G) and species abundance heatmaps (Fig. 7H) consistently demonstrated that the microbial profile of the AbLCP-treated mice was more analogous to that of healthy controls and distinctly different from the diseased model group. In contrast, the microbial profiles of PMB-, CM13-, and LCP-treated mice showed only partial shifts toward the healthy state, with CM13 exhibiting the greatest improvement among the three, followed by LCP and PMB (fig. S22, D to F). Collectively, AbLCP therapy restores a dysbiotic gut microbiota to a state structurally and functionally aligned with a healthy microbial ecosystem.

Mechanism of the superior therapeutic values

Beyond direct bactericidal action, AbLCP exhibited potent immunomodulatory functions by neutralizing LPS. This neutralization capacity directly translated to a profound suppression of LPS-induced inflammatory responses in RAW264.7 macrophages. Pretreatment with AbLCP markedly reduced the secretion of key pro-inflammatory cytokines, including IL-6 and IL-1β (Fig. 8, A and B). Furthermore, LPS is known to potently stimulate macrophages to produce nitric oxide (NO) via the induction of inducible nitric oxide synthase(iNOS). AbLCP treatment effectively blocked this pathway, reducing NO production by ∼92% compared to macrophages treated with wild-type M13 phages (Fig. 8C). Flow cytometry confirmed that AbLCP attenuated LPS-driven M1 polarization, decreasing CD86⁺ macrophages to baseline levels (Fig. 8, D and E). These data demonstrate that by efficiently neutralizing LPS, AbLCP attenuated its ability to hyperactivate macrophages, thereby curbing the release of inflammatory mediators and modulating the host immune response toward a less inflammatory state.

Fig. 8. Immunomodulatory and antimicrobial mechanisms underlying the enhanced therapeutic efficacy of AbLCP.

(A to C) AbLCP suppresses LPS-induced inflammatory activation in RAW264.7 macrophages. (A and B) Secretion of pro-inflammatory cytokines (A) IL-6 and (B) IL-1β after LPS stimulation with or without pretreatment. (C) Production of NO. (D and E) AbLCP mitigates LPS-driven M1 macrophage polarization. (D) Representative flow cytometry histograms showing the expression of the M1 marker CD86 (FITC) across treatment groups: DMEM (control), LPS, M13, CM13, LCP, and AbLCP. (E) Quantification of the MFI of CD86⁺. (F to I) Proteomic analysis reveals distinct protein expression profiles. (F) Bar graph showing the number of up-regulated and down-regulated proteins in P. aeruginosa treated with CM13 or AbLCP versus control. (G) Scatterplot depicting the expression fold change and statistical significance of differentially produced proteins. (H) Heatmap visualizing the expression patterns of all significantly differentially produced proteins across treatment groups. (I) PCoA plot showing the global proteomic profile separation between M13- and AbLCP-treated bacteria. (J and K) Validation of key stress response proteins, the up-regulation of (J) TolB and (K) Pal, components of the Tol-Pal system involved in outer membrane integrity, in AbLCP-treated P. aeruginosa. Data are presented as the means ± SEM (n = 3 or 5). Statistical analyses were performed using a one-way ANOVA followed by Tukey’s post hoc test. **P < 0.01; ****P < 0.0001; ns, not significant, P > 0.05.

Proteomic analysis of P. aeruginosa treated with AbLCP versus naked CM13 in simulated intestinal fluid revealed 22 differentially produced proteins, with 11 up-regulated and 11 down-regulated in the AbLCP group (Fig. 8, F to I). Notably, the up-regulated proteins delineated a multifaceted bacterial stress response. This included the marked up-regulation of TolB and Pal, core components of the Tol-Pal system essential for outer membrane integrity (Fig. 8, J and K). This finding directly indicates that bacteria attempt to reinforce their outer membrane against the combined assault of PMB and phage binding, a clear membrane damage alarm triggered by the dual-targeting construct. Concurrently, significant up-regulation of stress proteases ClpP2 and CtpA (fig. S23, A and B) points to intensive protein quality control and damage repair efforts. Furthermore, up-regulation of the motility proteins FliC (flagellin) and PilA (pilin) (fig. S23, C and D) reflects escape behavior. This multifaceted pressure, compromising membrane integrity, inducing overwhelming repair, and provoking escape, collectively underpins the enhanced bactericidal efficacy of AbLCP.

DISCUSSION

We introduce AbLCP, a double-targeting nanophage platform that successfully merges the genetic programmability of bacteriophages with the delivery capacity of LNPs. This design addresses key translational barriers in phage therapy including environmental instability, poor biofilm penetration, inability to target intracellular bacteria, and rapid clearance. The dual-targeting strategy creates a synergistic, multivalent binding effect that enhances specificity and avidity for Gram-negative pathogens. This results in superior functional performance, enabling AbLCP to penetrate complex biofilm matrices more effectively than single-component systems. The LNP shell protects the phage during oral delivery, promotes endosomal escape, and acts as a stimulus-responsive depot that releases phage cargo upon exposure to specific bacterial enzymes at the infection site.

By facilitating cellular uptake and endosomal escape, the AbLCP platform transforms the traditionally extracellular M13 phage into an effective intracellular therapeutic. It addresses a critical clinical need, as intracellular reservoirs are often responsible for chronic and recurrent infections. Proteomic analysis provides a compelling molecular narrative for this efficacy, showing that AbLCP induces a profound cellular stress response in bacteria, indicative of a more potent and comprehensive antibacterial assault compared to naked phages. In vivo studies demonstrated the platform versatility via both oral (gastrointestinal) and systemic (peritonitis and hindlimb) routes. AbLCP achieved marked bacterial reduction, resolved inflammation, restored gut microbiota diversity, and accumulated precisely at infection sites. Notably, PMB required a more than 500- or 1000-fold higher dose to match the efficacy of AbLCP in the peritonitis or enteritis model, respectively, underscoring the efficiency of this targeted approach. Furthermore, hematoxylin and eosin (H&E) staining revealed that major organs from AbLCP-treated mice appeared nearly normal, comparable to those of healthy controls (fig. S24A). The hematological analysis revealed that AbLCP treatment, even at efficacious doses, did not induce significant abnormalities in key blood parameters (white blood cells, lymphocytes, neutrophils, red blood cells, hemoglobin, and platelets) compared to controls (fig. S24, B to G). The body weight dynamics during the treatment period further confirmed the systemic safety of this platform (fig. S25).

This dual-targeting system has important advantages over conventional phage therapy, which mainly uses lytic phages (Fig. 9). First, phage therapy faces a pivotal translational barrier of insufficient in vivo stability and off-target accumulation. The AbLCP system shields phage particles from degradation and immune elimination. Second, the dual-targeting LNP system enables AbLCP to efficiently penetrate into dense biofilms and precisely capture extracellular and intracellular Gram-negative bacteria. These are critical bottlenecks of conventional phage therapy. Moreover, instead of using lytic phage, this system avoids massive abrupt bacterial lysis and can neutralize the released LPS by the antibodies, thereby alleviating host inflammatory responses. Third, progressive bacterial resistance evolution remains a universal threat to any antibacterial strategy, including the AbLCP system. Nevertheless, the phage is only used for bacterial capture, and cell death is caused by the conjugated PMB instead of the phage itself. Thus, the downstream events including infection and replication are not relevant, cutting down the emergence probability of phage-resistant bacteria. Fourth, the uncontrollable self-replication dynamics make the in vivo pharmacokinetic behaviors of native phages unpredictable. The burst release of endotoxins may trigger a harmful immune response, which can be lethal for immunocompromised and vulnerable patients. The AbLCP particle is nonreplicative. Therefore, its pharmacokinetic and pharmacodynamic profiles resemble standardized antibacterial drugs rather than replicative living microorganisms. This design simplifies clinical dosage regulation, facilitating practical clinical translation of phage therapy. The phage component can be reengineered to display different targeting ligands or to carry different cargoes. The LNP composition can be tuned for different release kinetics or to target specific tissues. This establishes AbLCP as a blueprint for a next-generation class of precision antimicrobials adaptable to a broad range of MDR pathogens. In summary, AbLCP integrates synthetic biology and nanomedicine to deliver a targeted, robust, and multifunctional therapeutic strategy with clear potential for clinical translation against recalcitrant Gram-negative infections.

Fig. 9. Schematic illustration of precision phage therapy, a dual-targeting LNP system for chimeric phage delivery to enhance biofilm penetration, intracellular bacterial inhibition, and antibacterial efficacy.

(A) AbLCP exhibits significantly deeper penetration into biofilms and is degraded by phospholipases (PLs) to release CM13. (B) AbLCP is capable of cellular internalization, endosomal escape, and clearance of intracellular bacterial infections, whereas naked CM13 cannot enter host cells. (C) AbLCP remains stable in harsh in vivo environments, attenuates inflammation, restores the microbiota via LPS neutralization, and selectively eliminates bacterial pathogens.

MATERIALS AND METHODS

Materials and strains

P. aeruginosa PAO1 strain [American Type Culture Collection (ATCC) 15692], K. pneumoniae (ATCC 700603), A. baumannii (ATCC 19606), and E. cloacae (ATCC 13047) were purchased from the ATCC (US). An mCherry-expressing P. aeruginosa PAO1 strain was constructed by transformation with the plasmid pBBR1MCS2-Tac-mCherry [Amp⁺, MiaoLing Plasmid Platform (China)]. HEK293 cells (FH0242), RAW264.7 macrophages (FH0328), and THP-1 cells (FH0112) were obtained from FUHENG BIOLOGY (China). Dulbecco’s modified Eagle’s medium (DMEM), RPMI-1640 medium, heat-inactivated fetal bovine serum, penicillin-streptomycin, and 0.25% trypsin-EDTA were purchased from Gibco (US). The pADL-10b vector is from Antibody Design Labs. Lipids including 1,2-dioleoyl-3-trimethylammonium-propane (DOTAP), 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC), cholesterol, 1,2-distearoyl-sn-glycero-3-phosphoethanolamine-N-[amino(polyethylene glycol)-2000] (DSPE-PEG2000), and DSPE-PEG2000-NHS were obtained from Guangzhou Tanshtech Co., Ltd. (China). The anti-His tag antibody [HRP (horseradish peroxidase)–conjugated] was purchased from Cell Signaling Technology (US). The PE-conjugated anti-M13 antibody (Mouse Monoclonal, 11973-MM05T-P) was purchased from Sino Biological (China). The fluorescein isothiocyanate (FITC) anti-mouse CD86 antibody (11-0862-81) was purchased from Invitrogen (US). Anti-Ly6G Rabbit pAb (GB11229) was obtained from Servicebio (China). The anti-Pseudomonas antibody (ab68538) was purchased from Abcam (UK). Goat Anti-Rabbit IgG H&L (Alexa Fluor 488, ab150077) was purchased from Abcam (UK). Cy3-conjugated Goat Anti-Rabbit IgG (H + L) (GB21303) was obtained from Servicebio (China). Enzyme-linked immunosorbent assay (ELISA) kits for mouse IL-6 (SLBA079MU), TNF-α (SLBA133MU), and CXCL1/KC were purchased from Lablead (China). Griess reagent for NO detection was obtained from Beyotime Biotechnology (China).

Construction of the chimeric phage

The anti-LPS single-chain variable fragment (scFv) was derived from the monoclonal antibody WN1 222-5 and displayed on the receptor binding protein of M13 phage to replace its N-terminal domain (49, 63). To construct the phagemid for phage display, the scFv gene was synthesized by Tsingke Biotechnology with homology arms added to both termini (CAGCCGGCCATGGCC at the 5′ end and GGCTCTGGTGGTGGT at the 3′ end). The pADL-10b vector backbone was amplified by polymerase chain reaction with primers Mid-F and Mid-R (table S1) to remove the N-terminal domain of the g3p protein. The linearized vector (100 ng) and the synthesized scFv insert (50 ng) were assembled using a homologous recombination method with the DNA Assembly Mix Plus kit (Lablead, D0204P) according to the manufacturer’s protocol. The reaction mixture contained 5 μl of 2×DNA Assembly Mix and was adjusted to a final volume of 10 μl with double-distilled water. The mixture was incubated at 50°C for 5 to 15 min. The E. coli DH5α competent cells were transformed with the assembled product, designated as the recombinant phagemid. Positive clones were confirmed by Sanger sequencing. The E. coli TG1 cells were transformed with the correct phagemid construct for phage production. When the culture reached the exponential growth phase, helper phage CM13D3 was added. The culture was induced under conditions of 25°C, kanamycin (50 μg/ml), and 200 μM isopropyl β-d-1-thiogalactopyranoside, with shaking at 250 r.p.m. for 8 to 16 hours to facilitate chimeric phage assembly and secretion. The supernatant was collected by centrifugation (4000g for 10 min), and 2.5 M NaCl/20% PEG-8000 (polyethylene glycol, molecular weight 800) (w/v) (one-fourth volume of the supernatant) was added to precipitate the phage overnight at 4°C. The PEG/NaCl precipitation step was repeated once to purify the phage, and the titer (PFU/ml) was determined. The purified phage (10¹³ PFU/ml) was incubated with PMB (10 mg/ml) and NHS (10 mg/ml) in PBS buffer (pH 7.2). EDC (10 mg) was added to the solution every 30 min for three times to efficiently activate the carboxyl groups from the phage. The reaction proceeded for another 2 hours after adding EDC, and the chimeric phage CM13 was purified by extensive dialysis against PBS at 4°C. The drug-to-virion ratio of the formed chimeric phage CM13 was quantified using an 1100 LC/MSD Trap system for two-dimensional ion trap LC-MS.

Construction of anti-LPS antibody

For soluble antibody production, the anti-LPS scFv gene was synthesized with a C-terminal 6×His tag and different homology arms (CAGCCGGCCATGGCC at the 5′ end and ACTAGTGGCCCGGGA at the 3′ end). The pADL-10b vector was linearized by polymerase chain reaction with primers Pure-F and Pure-R (table S1). The linearized vector (100 ng) and the synthesized insert (50 ng) were assembled using the same homologous recombination method described above. The resulting plasmid encoded the anti-LPS scFv fused with a 6×His tag, preceded by a PelB signal peptide for periplasmic expression. E. coli BL21(DE3) was transformed with the construct, which was further confirmed by Sanger sequencing. Then, a single colony with a verified plasmid was inoculated into LB medium containing ampicillin (100 μg/ml) and cultured overnight at 37°C, shaking at 220 rpm. The overnight culture was diluted 1:100 into fresh LB medium and grown at 37°C until the optical density at 600 nm reached 0.6 to 0.8. Protein production was induced by adding isopropyl β-d-1-thiogalactopyranoside to a final concentration of 0.5 mM, followed by continued incubation at 25°C for 16 hours with shaking at 220 rpm. Bacterial cells were harvested by centrifugation at 8000g for 20 min at 4°C. The supernatant was collected and filtered through a 0.45-μm membrane. The clarified supernatant was loaded onto a Ni NTA Beads 6FF column (Lablead, N30210) preequilibrated with binding buffer [20 mM sodium phosphate, 500 mM NaCl, and 20 mM imidazole (pH 7.4)]. The column was washed with 10 column volumes of binding buffer, and bound proteins were eluted with a linear gradient of imidazole (20 to 500 mM) in the same buffer using an ÄKTA pure chromatography system (Cytiva). Elution fractions containing the target protein were pooled and concentrated using Amicon Ultra-15 centrifugal filters (10-kDa molecular weight cutoff, Millipore). The concentrated sample was further purified by size exclusion chromatography on a Superdex 200 Increase 10/300 GL column (Cytiva) equilibrated with PBS (pH 7.4). Peak fractions were collected, and the identity of the purified antibody was confirmed by Western blotting using an anti-His tag antibody (Cell Signaling Technology, 9991). The purified antibody was stored at −80°C until further use. The primers and gene sequences used for antibody and chimeric phage construction are listed in table S1 and text S1.

Construction of AbLCP

LNPs were formulated with the following weight percentages: 40% DOTAP, 35% DOPC, 20% cholesterol, and 5% DSPE-PEG2000 or DSPE-PEG2000-NHS. Stock solutions of each lipid were prepared at 100 mM in chloroform. For a typical 200-μl formulation, 11 μl of DOTAP, 9.6 μl of DOPC, 5.48 μl of cholesterol, and 1.1 μl of DSPE-PEG2000 or DSPE-PEG2000-NHS were mixed in a 2-ml glass vial, followed by addition of 155.5 μl of chloroform and 17.3 μl of methanol. The mixture was vortexed thoroughly, and the organic solvents were evaporated under a gentle stream of argon gas to form a thin lipid film on the vial wall. The film was further dried in a vacuum desiccator for 2 hours to remove residual solvent. The lipid film was rehydrated with phage suspension (CM13 at 1 × 10¹³ PFU/ml) and vortexed to form a crude liposome suspension. The suspension was extruded 33 times through a 200-nm polycarbonate membrane (Avanti, 610006) using a miniextruder (Avanti, A81411) to obtain uniform LCPs on a plate heater at 37°C. The LCP was purified by size exclusion chromatography (Superdex 200 Increase column, Cytiva). For AbLCP preparation, LCP containing DSPE-PEG2000-NHS was incubated with the purified anti-LPS antibody at 4°C for 4 hours. AbLCPs were then purified by size exclusion chromatography (Superdex 200 Increase column, Cytiva) to remove unbound antibodies. AbLCPs were characterized by AFM (Bruker MultiMode 8), Zetasizer (Malvern Zetasizer Nano ZS), cryo-TEM (Thermo Fisher Scientific Talos Arctica), and SIM (CSR Biotech Co., Ltd., China). The PMB concentration of AbLCP was determined by LC-MS (1100 LC/MSD Trap, Agilent) after rupturing the particles with 0.6% Triton X-100 for 30 min at room temperature.

AFM imaging

Ten microliters of the diluted sample (CM13 or AbLCP, 10⁹ PFU/ml or equivalent) in ultrapure water was deposited onto a freshly cleaved mica and allowed to adsorb for 2 min. The sample was dried thoroughly under a lamp before AFM imaging in air. AFM imaging was performed on a Bruker MultiMode 8 atomic force microscope using a ScanAsyst-Air probe (model: SCANASYST-AIR) mounted on an air probe holder. Scanning parameters were set with the following initial values: scan size = 1 μm; X offset and Y offset = 0; scan angle = 0; scan rate = 0.977 Hz; sample/line = 256. After engagement, the scan size was gradually adjusted to the target range to acquire optimal images. All AFM images were processed and analyzed using NanoScope Analysis 1.9 software.

Zeta potential and DLS measurements

The zeta potential and hydrodynamic size of AbLCP were measured by Zetasizer (Malvern, ZS90) with an equilibration time of 120 s. Seven hundred fifty microliters of the sample was transferred to a cuvette (Malvern, DTS1060) for zeta potential measurement, and 1 ml of the sample was transferred to a disposable polystyrene cuvette (Malvern, DTS0012) for size measurement. Each measurement was repeated three times.

SIM imaging

SIM imaging was performed using a HIS-SIM (High Intelligent and Sensitive SIM) system (CSR Biotech Co., Ltd., China). A 200-μl aliquot of the AbLCP sample was placed into a confocal dish. Imaging was performed using a 100× oil immersion objective. Fluorescence images were acquired with 488- and 638-nm laser lines, together with bright-field imaging. The imaging mode was set to 2D-SIM-1, and the exposure time was set to 200 ms. SIM image reconstruction was performed using the SimAna software provided by the manufacturer.

Cryo-TEM imaging

A 3-μl aliquot of the AbLCP sample (∼10¹¹ particles/ml) was applied onto a holey carbon-coated copper grid (Quantifoil R 2/1/3, 100 Holey Carbon Film, Cu 300 mesh, Quantifoil Micro Tools GmbH, Germany). The grid was blotted for 2 to 4 s under 100% humidity at 22°C and immediately plunge-frozen in liquid ethane cooled by liquid nitrogen using a Vitrobot (Thermo Fisher Scientific). The vitrified grid was transferred to a cryoholder and maintained at ∼−180°C during imaging. Cryo-TEM images were acquired using a Krios G4 transmission electron microscope (Thermo Fisher Scientific) operated at a 300-kV accelerating voltage. Images were recorded using a Ceta 16M complementary metal-oxide semiconductor camera (Thermo Fisher Scientific) at nominal magnifications of 1500× to 30,000×.

SDS-PAGE and Western blotting

Protein samples were prepared by mixing 90 μl of each sample (empty LNP, LCP, AbLCP, or free anti-LPS antibody) with 30 μl of 4× reducing SDS-PAGE loading buffer containing dithiothreitol (Lablead, G2526). The mixtures were heated at 100°C for 5 min and cooled to room temperature. Samples were loaded onto 10% precast SDS-PAGE gels (Lablead, P01012) and electrophoresed at a 150-V constant voltage for 60 min. For Coomassie Blue staining, gels were incubated in Coomassie Brilliant Blue staining solution (Tsingke Biotechnology, DLW601) for 20 min and destained with distilled water for 1 hour before imaging. For Western blot analysis, proteins were transferred from the SDS-PAGE gel onto a polyvinylidene difluoride membrane (Lablead, PM0453) using a semidry transfer system at a 110-V constant voltage for 100 min. The membrane was blocked with 5% bovine serum albumin in TBST (tris-buffered saline with 0.1% Tween 20) at room temperature for 1 hour. The blocked membrane was incubated with an anti–His-HRP antibody (Cell Signaling Technology, 9991) at 4°C overnight. After triplicate 10-min washing with TBST, the membrane was incubated with 500 μl of chemiluminescent substrate (Millipore) for 1 min and imaged using a chemiluminescence imaging system (1600, Tanon).

Stability test of AbLCP

The zeta potential of AbLCP was measured in PBS (pH 7.4) by Zetasizer (Malvern Zetasizer Nano ZS). pH stability was assessed by incubating particles in buffers ranging from pH 2.5 to 11.5 for 2 hours at 37°C before DLS measurement. Phospholipase C responsiveness was tested by incubating AbLCP with phospholipase C (0 to 1 U/ml) in tris-HCl buffer (pH 7.4) at 37°C. Size changes were monitored by DLS at 1 and 4 hours. The long-term stability of AbLCP was evaluated by measuring the zeta potential, size, and antibacterial efficiency over 28 days of storage at 4°C.

Binding affinity assays

Affinity experiments were performed on the WeSPR300 instrument (Wellspring, China) in PBS with Tween 20 at room temperature. LPS or whole bacterial cells were immobilized on the surface of a 96-well gold sensor chip. Analyte solutions containing M13, CM13, and AbLCP protein at various concentrations were introduced into the detection wells, and real-time binding and dissociation signals were recorded at a frame rate of 1 fps. For each concentration, association was monitored until equilibrium was approached, followed by buffer exchange to record the dissociation phase. Binding curves were processed using WeSPR300 analysis software. Signals were extracted as the mean red-channel intensity within the defined region of interest. Kinetic parameters (K_d) were derived by global fitting of association and dissociation phases (79, 80).

Bacterial capture capability evaluation by flow cytometry

A single colony of P. aeruginosa, K. pneumoniae, E. cloacae, or A. baumannii was inoculated in LB broth and grown overnight. The cultures were diluted 1:100 into fresh LB medium and incubated at 37°C with shaking until the exponential growth phase was reached [OD₆₀₀ (optical density at 600 nm) ≈ 0.4 to 0.6]. Bacterial concentrations were determined by OD₆₀₀ calibration. Approximately 1 × 10⁷ CFUs of each bacterial strain were harvested by centrifugation at 5000g for 5 min and washed twice with 1 ml of PBS. The bacterial pellets were resuspended and incubated with 1 ml of either PBS (negative control), free PE dye (200 ng/ml), PE-labeled CM13 (1 × 10¹⁰ PFU/ml), or PE-labeled AbLCP (1 × 10¹⁰ PFU/ml equivalent) in the dark with rotation at room temperature for 2 hours. After incubation, the samples were washed three times with 1 ml of PBS for each to remove unbound reagents, resuspended in 1 ml of PBS, and transferred to flow cytometry tubes. The samples were analyzed on a CytoFlex flow cytometer (Beckman, CytoFlex). Data were analyzed using FlowJo software (version 10.8).

Antibacterial activity against planktonic bacteria

Time-kill assays were conducted with planktonic bacteria at an initial concentration of 1 × 10⁶ CFUs/ml. This bacterial density was selected because it approximates the physiological bacterial load in mild to moderate bacterial infections and early-stage infection in vivo. Such a moderate inoculum keeps bacteria in a normal physiological state without rapid nutrient exhaustion, supporting reliable monitoring of dynamic changes in viable bacteria over the 12-hour incubation period. Meanwhile, this starting concentration avoids component saturation caused by excessively high bacterial titers, allowing an accurate comparison of the bactericidal kinetics and efficacy among PBS control, free PMB, CM13, and AbLCP groups. Bacteria were treated with PBS (control), PMB, CM13, or AbLCP. The free PMB concentrations for P. aeruginosa PAO1, K. pneumoniae, A. baumannii, and E. cloacae in the assays are 2, 2, 16, and 250 μg/ml, respectively. The equivalent PMB concentrations of CM13 or AbLCP for P. aeruginosa PAO1, K. pneumoniae, A. baumannii, and E. cloacae in the assays are 0.01, 0.04, 0.2, and 6 μg/ml, respectively. Aliquots were taken at 0, 0.5, 1, 2, 3, 6, 9, and 12 hours; serially diluted; plated on LB agar; incubated overnight; and enumerated for CFUs.

Bacterial eradication in biofilm

Bacterial biofilm formation was assessed in 96-well plates following incubation of bacteria at 30°C for 48 hours, with the medium replenished every 24 hours. Planktonic cells were gently removed by aspiration, and the wells were washed with 0.1 M PBS (pH 7.2). PMB, CM13, or AbLCP was added, and the plates were incubated for 12 hours. The free PMB concentrations for P. aeruginosa PAO1, K. pneumoniae, A. baumannii, and E. cloacae in the assays are 20, 20, 160, and 2.5 mg/ml, respectively. The equivalent PMB concentrations of CM13 or AbLCP for P. aeruginosa PAO1, K. pneumoniae, A. baumannii, and E. cloacae in the assays are 0.1, 0.4, 2, and 60 μg/ml, respectively. Metabolically active cells were determined using the XTT {sodium 3′-[1-(phenylaminocarbonyl)-3,4-tetrazolium]-bis(4-methoxy-6-nitro) benzenesulfonic acid hydrate} assay (81). An XTT-menadione solution (XTT: 1 μg/μl; menadione: 0.4 mM; mixed 1:1, v/v) was added. The plates were then incubated statically at 30°C for 5 hours in the dark, and absorbance was measured at 490 nm.

Biofilm penetration assay

P. aeruginosa biofilms were grown in confocal dishes (NEST, 1050001) for 48 hours. Biofilms were treated with PE-labeled phages for 0.5 hours, washed, and imaged by confocal laser scanning microscopy (CLSM; FV3000) with Z-stack acquisition. Maximum penetration depth and three-dimensional distribution were analyzed using ImageJ software. For quantitative transport assessment, biofilms were grown on 0.4-μm polyester membranes (Corning) for 48 hours. Phages were added to the apical chamber of a transwell system. After 4 hours, samples from the basal chamber were collected, and phage titers were determined by the double-layer agar assay.

Cellular uptake assay

HEK293 cells or THP-1–derived macrophages (5 × 10⁴ cells per well) were incubated with PE-labeled AbLCP or naked CM13 (equivalent to 10⁹ PFU/ml) for 1, 2, and 4 hours. For confocal imaging, cells were stained with Hoechst 33342 (Beyotime Biotechnology, C1027) for 10 min before the end of incubation, washed three times with PBS, and imaged in 500 μl of PBS using CLSM. For flow cytometry, cells were detached with 200 μl of trypsin, incubated at 37°C for 5 min, quenched with 600 μl of PBS, centrifuged (500g, 4°C), washed three times with PBS, and analyzed using a flow cytometer (Beckman, CytoFlex).

Lysosomal escape assay

HEK293 cells were cocultured with PE-labeled AbLCP or naked CM13 for 1 or 2 hours. Ten minutes before observation, cells were stained with Hoechst 33342 and LysoTracker Green (Beyotime Biotechnology, C1047S). Cells were washed three times with PBS before CLSM imaging. Excitation wavelengths were set at 405, 504, and 562 nm.

Intracellular bactericidal activity

HEK293 cells (5 × 10⁴ cells per well) were infected with mCherry-expressing P. aeruginosa at a multiplicity of infection of 50 for 12 hours. Extracellular bacteria were killed by carbenicillin treatment for 2 hours. Cells were then treated for 4 hours with the following agents: blank medium, PBS, PMB, CM13, LCP, or AbLCP. Phages were labeled with PE as described and used at a concentration of 10⁹ PFU/ml. After 4 hours, the supernatant was replaced with PBS, and cells were imaged by confocal microscopy. Cells were lysed with 0.1% Triton X-100, and lysates were plated for CFU counting. Bacterial load was also assessed by flow cytometry on the basis of mCherry fluorescence.

Mouse studies

All mouse experiments conducted in this work complied with national and institutional regulations on laboratory animal welfare. The entire animal study protocol has obtained formal ethical approval from the Institutional Animal Care and Use Committee (IACUC) of Huazhong University of Science and Technology. The official experimental animal ethics approval number is HUST-IACUC-2026-0038. Male Balb/c mice (6 to 8 weeks old) were obtained from Wuhan Hualianke Biotechnology Co., Ltd. Mice were randomly assigned to groups and housed under standard conditions (12-hour light/dark cycle; 22 ± 2°C; 55 ± 10% humidity) with free access to water and standard chow.

In vivo targeting and fluorescence imaging of AbLCP

Male Balb/c mice (6 to 8 weeks old, n = 6) were shaved before the experiment. A hindlimb bacterial infection model was established by intramuscular injection of 100 μl of PBS into the left leg and 100 μl of P. aeruginosa (5 × 10⁷ CFUs/ml) into the right leg. Cy7-labeled AbLCP (1 × 10¹⁰ PFU) was administered via tail vein injection. Whole-body fluorescence imaging was performed at 0, 1, 3, 6, 9, 12, and 24 hours using an in vivo imaging system (IVIS; AniView 600 Phoenix X). At specific time points, mice were euthanized, and the heart, liver, spleen, lung, kidney, left leg muscle, and right leg muscle were excised for ex vivo fluorescence intensity assessment.

Treatment of P. aeruginosa–induced peritonitis

Mice were intraperitoneally injected with P. aeruginosa (5 × 10⁷ CFUs per mouse in 100 μl). After 24 hours, mice were randomly divided into six groups (n = 5 per group) and treated by daily intraperitoneal injection with PBS, PMB (70 μg per dose), CM13 (1.34 × 10¹⁰ PFU per dose, equivalent to 0.67 μg of PMB per dose), LCP (equivalent to 0.134 μg of PMB per dose), or AbLCP (equivalent to 0.134 μg of PMB per dose). Last, mice were injected intraperitoneally with 2 ml of cold Hanks’ balanced salt solution and then euthanized. Peritoneal lavage fluid was collected to determine total, extracellular, and intracellular CFUs. One-third of the fluid was used for total CFU quantification. One-third was centrifuged, and the supernatant was used for extracellular CFU quantification. One-third was treated with a mixture of Hanks’ balanced salt solution containing 0.1% bovine serum albumin and 0.1% Triton X-100 to quantify intracellular CFUs. Blood was collected from the orbital sinus to measure serum levels of IL-6 and CXCL1/KC. Major organs were harvested for histology using H&E staining. Tissue immunofluorescence was performed to image bacterial content in the liver, spleen, and kidney.

Treatment of P. aeruginosa–induced enteritis

A model of P. aeruginosa intestinal infection was established in male Balb/c mice (6 to 8 weeks old). Mice were pretreated with an antibiotic cocktail in drinking water for 7 days, followed by oral infection with 5 × 10⁸ CFUs of P. aeruginosa. Starting 24 hours postinfection, mice were orally gavaged daily for 9 days with PBS, PMB (0.5, 70, and 140 μg per dose), CM13 (equivalent to 0.134, 0.335, 0.67, and 1.34 μg of PMB per dose), LCP (equivalent to 0.134, 0.335, 0.67, and 1.34 μg of PMB per dose), or AbLCP (equivalent to 0.134, 0.335, 0.67, and 1.34 μg of PMB per dose). Mice were weighed daily. All mice were euthanized 24 hours after the last dose. Bacterial loads in blood, ileum, and colon were measured. Intestinal tissues were collected for histology (H&E), immunofluorescence (anti-Ly6G for neutrophils), and ex vivo fluorescence imaging. Plasma was separated from whole blood by centrifugation (2000g, 15 min, 4°C). Levels of IL-6 and TNF-α were measured using ELISA kits (Lablead, SLBA079MU and SLBA133MU).

Gut microbiota analysis

Fecal samples were collected at the end point of the enteritis model. Genomic DNA was extracted using the DNeasy PowerSoil Pro Kit (Qiagen). The V3-V4 region of the 16S ribosomal RNA gene was amplified with primers 341F/806R and sequenced on an Illumina MiSeq platform (2 × 300 base pairs). Raw reads were processed using the DADA2 pipeline in QIIME2 (version 2023.5) to infer ASVs. Taxonomy was assigned against the SILVA 138 database. α diversity (Shannon index) and β diversity (Bray-Curtis dissimilarity and PCoA) were calculated. Differential abundance and shared ASVs (Venn diagram) were analyzed. UPGMA clustering and heatmaps were generated in R (version 4.3.1) using the phyloseq package.

Proteomic analysis

P. aeruginosa PAO1 was treated with CM13 or AbLCP (equivalent to 10⁹ PFU/ml CM13) in simulated intestinal fluid for 4 hours. Bacterial cells were collected, lysed, and digested with trypsin. Peptides were analyzed by LC–tandem MS on a timsTOF Pro 2 (Bruker) coupled with a nanoElute LC system. Data were processed using MaxQuant (version 2.4.8.0) against the P. aeruginosa PAO1 UniProt database. Differential expression analysis was performed with Perseus (version 2.0.11.0) using thresholds of |log₂ fold change| > 1 and adjusted P < 0.05.

Immunomodulation assay

RAW264.7 macrophages were pretreated with AbLCP, LCP, and CM13 (10⁸ PFU/ml equivalent) for 1 hour and then stimulated with LPS (100 ng/ml, E. coli O111:B4) for 24 hours. Culture supernatants were assayed for IL-6, IL-1β, and TNF-α by ELISA and for NO using the Griess reagent. For polarization analysis, cells were stained with an FITC anti-mouse CD86 antibody and analyzed by flow cytometry.

Histopathology

Major organs and tissues were collected from mice, rinsed with PBS, longitudinally dissected, and fixed overnight in formalin before paraffin embedding. Paraffin sections (5 μm thick) were stained using an H&E staining kit (Beyotime Biotechnology, C0105M). After deparaffinization and dehydration, slides were treated with saffron staining solution for 10 min, rinsed with running water for 10 min, placed in a differentiation solution for 1 min, rinsed again for 10 min, and stained with eosin solution for 2 min. To measure mucus layer thickness, 5-μm-thick sections were stained using an AB-PAS staining kit (MXB Biotechnologies, MST-8005). Tissue sections were stained with Alcian blue (pH 2.5) solution for 10 min, incubated in acetic acid solution for 2 min, rinsed with running water for 5 min, treated with saffron solution for 10 min, stained with Schiff’s solution in the dark for 15 min, and lastly treated with sodium bisulfite for 1 min. Sections were mounted with neutral resin and imaged using a digital pathology slide scanner (Leica Biosystems Aperio CS2).

Hematology and safety assessment

For hematological analysis, whole blood was collected in EDTA tubes and analyzed using a veterinary hematology analyzer (BC-2800vet) for complete blood count parameters, including white blood cells, lymphocytes, neutrophils, red blood cells, hemoglobin, and platelets.

Statistical analysis

Statistical comparisons between two independent groups were performed using an unpaired two-tailed t test or multiple t tests. Statistical analyses were performed using a one-way analysis of variance (ANOVA) followed by Tukey’s post hoc test using GraphPad Prism 8. P < 0.05 was considered statistically significant. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns, not significant, P > 0.05.

Supplementary Materials

The PDF file includes:

Figs. S1 to S25

Table S1

Text S1

Legends for data S1 and S2

Other Supplementary Material for this manuscript includes the following:

Data S1 and S2