Revolutionizing antibiotic therapy: Polymyxin B and Fe²⁺-enriched liposomal carrier harness novel bacterial ferroptosis mechanism to combat resistant infections

Abstract

To address the pressing issue of bacterial resistance, antibiotics with new mechanisms were urgently needed; yet, the majority of efforts centered on discovering novel structural compounds, often plagued by lengthy research timelines and unpredictability. In this study, we introduce an alternative strategy that rejuvenates outdated antibiotics through a unique delivery system. Specifically, we leveraged polymyxin B (PMB) and created a liposomal carrier encapsulating PMB and Fe²⁺, designated P/Fe@L-P. When administered to PMB-resistant Acinetobacter baumannii, P/Fe@L-P triggered a downregulation of Nrf2 and GPX4 proteins, accompanied by a significant surge in reactive oxygen species and malondialdehyde levels, signifying the induction of ferroptosis. This mechanism imparted potent antibacterial activity, with P/Fe@L-P achieving minimal inhibitory and bactericidal concentrations of 54 and 192 μM, respectively, outperforming free PMB (72 and 768 μM). In vivo evaluations in mice models further validated the superior efficacy of P/Fe@L-P over PMB in treating PMB-resistant Acinetobacter baumannii pneumonia. This work establishes a highly effective and practical “old drug, new trick” paradigm, potentially expediting the fight against the escalating threat of bacterial resistance.

Graphical abstract

Highlights

• •Liposomal co-delivery of polymyxin B & Fe²⁺ altered polymyxin B's bacterial interaction.
• •This liposome triggered ferroptosis, a novel bactericidal mode, in Acinetobacter baumannii.
• •The novel bactericidal mechanism enhanced efficacy against PMB-resistant Acinetobacter baumannii.

1. Introduction

The pressing concern of bacterial resistance, fueled by the misuse of antibiotics, has risen to the forefront of global public health, demanding urgent and innovative solutions. Notably, Acinetobacter baumannii (A. baumannii), a pathogen known to inflict various respiratory tract infections [1,2], has demonstrated an alarming increase in resistance to a broad range of antibiotics, including penicillin, carbapenems, and sulfonamides. This trend has facilitated the widespread dissemination of resistant A. baumannii strains.

In response to the growing threat of multidrug resistance in A. baumannii within clinical settings, polymyxin B (PMB) [3], once hailed as the “last line of defense”, has been reinstated for clinical use [4]. However, the scientific community now faces a significant challenge: the emergence of PMB-resistant A. baumannii strains. To combat this, clinicians have resorted to complex antibiotic combination therapies, such as the synergistic use of cefoperazone sulbactam sodium with imipenem/amikacin [5]. Unfortunately, despite these strategies, the pathogen has demonstrated notable resistance to certain antibiotic combinations, highlighting the limitations of relying solely on altering antibiotic regimens to outpace resistance. The relentless evolution of resistance often outpaces the development of new biotics, necessitating a shift in strategy. Therefore, a critical approach to addressing the global challenge of emerging drug-resistant bacteria lies in the exploration and development of antibiotics with novel bactericidal mechanisms. This approach offers a promising avenue to overcome the limitations of existing therapies and safeguard public health against the escalating threat of bacterial resistance.

Given the diverse array of mechanisms that confer drug resistance, one pivotal factor is the modification of the drug's target. As a result, significant efforts have been directed towards the development of drugs with novel chemical structures that target new sites or employ unprecedented mechanisms of action, thereby circumventing existing resistance mechanisms. Although some novel promising compounds with new action models have entered clinical trials, such as gepotidacin, zoliflodacin, ibezapolstat, MGB-BP-3, CRS-3123, afabicin and TXA-709 [6], this process is fraught with challenges. Complex and costly trial procedures, along with unpredictable variables during the trials, have collectively posed obstacles [7], significantly prolonging the journey from laboratory research to clinical practice. Hence, humanity faces an escalating and formidable challenge posed by microbial resistance, with the key task being the design of strategies to quickly and cost-effectively shorten the antibiotic discovery and development cycle. At the same time, there is an urgent need to accelerate the transition of these antibiotics from the laboratory to clinical practice, ensuring their timely and effective deployment.

Considering the long process from discovery to marketing of compounds with novel chemical structures, we introduce a strategy: for older antibiotics that have succumbed to resistance, we aim to revitalize their efficacy by optimizing drug delivery systems and reshaping their interaction with bacteria, so as to trigger novel bactericidal mechanisms. Our focus lies on PMB, a key last resort in the battle against infections [8], which uses its positive charge to bind with lipopolysaccharide (LPS) on Gram-negative bacteria, causing bacterial lysis and death [9,10]. However, this potent mechanism is hindered by significant nephrotoxicity [11], limiting its widespread clinical application. Compounding this issue, bacteria evade PMB's action by downregulating LPS expression, leading to drug resistance and making clinical management difficult [12].

To overcome this problem, we have developed an approach that moves beyond the conventional LPS-dependent bactericidal mechanism and focuses on a new bactericidal pathway based on ferroptosis. In eukaryotic systems, PMB can enhance the generation of reactive oxygen species (ROS) through the Fenton reaction [[13], [14], [15]], triggering the lipid peroxidation and ultimately ferroptosis [16,17]. By harnessing the catalytic capacity of Fe²⁺ in facilitating the Fenton reaction, we propose a strategy for the co-delivery of PMB and Fe²⁺ into the bacterial via liposomes. Despite the important role of iron acquisition and homeostasis in the virulence of various microbial pathogens [18,19], there is little research on ferroptosis in A. baumannii, leaving much to be explored. In this study, we aim to investigate whether the co-delivery of PMB and Fe²⁺ within bacterial cells can induce ferroptosis and thereby enhance bactericidal effects.

Liposomes, nanoscale vehicles made from phospholipids and cholesterol, have a membrane structure that closely resembles bacterial outer membranes [[20], [21], [22]], conferring them with exceptional drug delivery potential. We expect that the strong binding affinity and penetration ability of nanoliposomes will facilitate the delivery of PMB and Fe²⁺ into specific bacteria. Furthermore, the encapsulating effect of liposomes is likely to reduce direct interactions between PMB's positive charge and host cells, thus lowering its tissue toxicity.

We aim to administer this liposomal system via inhalation as a treatment for pneumonia caused by PMB-resistant A. baumannii, a pathogen that threatens the health of hospitalized individuals, particularly those in the intensive care units [23], requiring targeted intervention. Given the abundant mucus in the airways of patients with pulmonary infections [24,25], it is important for liposomes to effectively penetrate this mucus layer. Previous research has shown that reducing liposome diameter and PEGylating their surface can improve their ability to penetrate mucus [26]. Therefore, we have refined the liposome design by applying PEGylation and reducing particle size.

In summary, as shown in Fig. 1, this study was designed to create a straight forward PEGylated liposome, encapsulating both PMB and Fe²⁺ (designated as P/Fe@L-P). This approach aims to induce ferroptosis in A. baumannii, replacing the conventional LPS-dependent bactericidal mechanism, which is vulnerable to drug resistance, with a new mechanism. Our strategy seeks not only to overcome the PMB resistance issue by using ferroptosis as a novel antimicrobial approach but also to provide a practical pathway for “old drug, new mechanism”. Ultimately, this work aims to offer new insights and viable solutions to address the growing problem of bacterial drug resistance, accelerating the development cycle of new antibiotics.

Fig. 1.

The preparation and the potential ferroptosis-based antibacterial mechanisms of P/Fe@L-P. PEG: polyethyleneglycol; PMB: polymyxin B; ROS: reactive oxygen species; P/Fe@L-P: PMB and Fe²⁺ loaded liposomes with PEGylation.

2. Materials and methods

2.1. Materials and bacterial strains

PMB was purchased from Macklin Biochemical Co., Ltd. (Shanghai, China). Dipalmitoyl phosphatidylcholine (DPPC) was obtained from Lipoid GmbH (Cologne, Germany). Cholesterol was sourced from AVT Pharmaceutical Co., Ltd. (Shanghai, China). 1,2-Distearoyl-sn-glycero-3-phosphoethanolamine-N-[methoxy(polyethylene glycol)-2000] (DSPE-mPEG₂₀₀₀) was purchased from Pengshuo Biotechnology Co., Ltd. (Shanghai, China). H₃PO4 and FeCl₂·4H₂O were procured from Sinopharm Chemical Reagent Co., Ltd. (Beijing, China). Rhodamine B isothiocyanate, fluorescein isothiocyanate, acetone, methyl blue and 2,2′-bipyridyl were procured from Merck Co., Ltd. (Darmstadt, Germany). GPX4 and Nrf2 polyclonal antibodies were procured from Elabscience Biotechnology Co. Ltd. (Wuhan, China). Horseradish peroxidase coupled secondary antibody was procured from Abcam Co., Ltd. (Shanghai, China).

A. baumannii ATCC19606 was provided by Xuzhou Yuanyao Pharmaceutical Technology Research Institute Co., Ltd. (Xuzhou, China), and was isolated from sputum samples of hospitalized pneumonia patients at the Affiliated Hospital of Guangdong Medical University (Guangzhou, China). The induction method for the PMB-resistant strain (ATCC19606R) is detailed in Section 2.3.

Multidrug-resistant Klebsiella pneumoniae (K. pneumoniae) and Staphylococcus aureus (S. aureus) were isolated from sputum samples of hospitalized pneumonia patients at the Affiliated Hospital of Xuzhou Medical University (Xuzhou, China). The antibiotic resistance profiles of these bacteria are presented in Tables S1 and S2.

2.2. Preparation and characterization of P/Fe@L-P

2.2.1. Preparation of P/Fe@L-P

P/Fe@L-P was prepared using the thin-film dispersion method. Briefly, the oil phase was prepared by dissolving DPPC (90 mg), cholesterol (45 mg), and DSPE-mPEG₂₀₀₀ (16.8 mg) in CH₂Cl₂ (8 mL), and then CH₂Cl₂ was removed by rotary evaporation, forming a thin film. Then, 1 mL of aqueous phase containing PMB (40 mg), FeCl₂·4H₂O (28 mg), and H₃PO₄ (0.06%, w/v) was then added for hydration, resulting in Fe/P@L-P. The unencapsulated PMB and Fe²⁺ were removed using the Pierce™ Crossflow Filtration Cassettes (30 kDa; Thermo Scientific, Waltham, MA, USA). In order to minimize the liposome volume for inhalation by the rats in the animal experiment, the liposome was concentrated using the Pierce™ Crossflow Filtration Cassettes until the concentration of PMB reached to 6.25 mg/mL.

Liposomes without Fe²⁺ (P@L-P), without both Fe²⁺ and surface PEGylation (P@L), and without PMB and surface PEGylation (Fe@L) were formulated from the certain components with the same method. The fluorescence labeled liposomes were formulated from rhodamine B isothiocyanate (RITC)-labeled cholesterol (namely RITC-cholesterol) and/or fluorescein isothiocyanate (FITC)-labeled PMB (namely FITC-PMB). RITC-cholesterol was prepared by mixing RITC (60 mg/mL) with cholesterol (38 mg/mL) in ethanol and reacting overnight. FITC-PMB was prepared by mixing FITC (4 mg/mL) with PMB (13 mg/mL) in phosphate buffer saline (PBS, pH 7.4) and incubating overnight.

2.2.2. Characterization of P/Fe@L-P

Particle size and zeta potential of P/Fe@L-P were measured using a Nano Brook Omni Particle Size Analyzer (Brookhaven Instruments Corporation, Holtsville, NY, USA). The structure and morphology of the liposomes were evaluated by transmission electron microscopy (TEM; Tecnai G2 Spirit Twin, FEI Company, Hillsboro, OR, USA). To enhance the resolution of liposomes against the background in TEM images, we incorporated phosphotungstic acid (10 μg/mL), a negative staining agent commonly used in TEM, into the aqueous phase during the hydration process of liposome preparation, allowing it to be encapsulated within the liposomes.

To quantify the total PMB in the liposomes, samples were demulsified by acetone (acetone:sample:H₂O = 2:1:3, v/v/v). The free PMB was separated from the encapsulated drug by ultrafiltration centrifugation at 5,000 rpm for 10 min with a cut-off molecular weight of 100 kDa. PMB was detected by high-performance liquid chromatography (HPLC), and Fe²⁺ was detected using a total iron assay kit (Huankai Microbial, Guangzhou, China). The loading efficiency (LE%) and encapsulation efficiency (EE%) of PMB and Fe²⁺ could be calculated.

To measure the in vitro release of PMB, P/Fe@L-P suspension samples were diluted 10-fold with 0.9% NaCl and oscillated at 50 rpm and 37 °C. At various time intervals, 0.5 mL samples were withdrawn, and the released PMB from P/Fe@L-P was separated by ultrafiltration and detected by HPLC. The dilution by 0.9% NaCl was performed to lower the concentration of the released PMB, and this approach aimed to mimic the release environment within the body, where the drug, once released, would be swiftly cleared from the system via pulmonary absorption.

To assess the physical stability of the constructs, P/Fe@L-P was kept at 25 and 40 °C, and their particle size and EE% were measured at different time intervals over 30 days.

2.3. Induction of the PMB resistant-A. baumannii (ATCC19606R)

A. baumannii ATCC19606 strain was cultured on mueller-hinton B (MHB) agar media containing PMB (1.5, 3, 6, 12, 24, 48, or 72 μM). Single colonies growing on the media with the highest concentration of PMB were selected, and the minimum inhibitory concentration (MIC) of PMB for these colonies was measured as described in Section 2.4. Subsequently, the strain was cultured on media with higher concentration of PMB. This process was repeated until the MIC exceeded 3 μM [27] to obtain a PMB-resistant strain (ATCC19606R).

2.4. In vitro antibacterial activity assessment

The Kirby-Bauer disk diffusion method was employed for the inhibition zone test. Briefly, under aseptic conditions, 0.25 mL of P/Fe@L-P (45 μM for ATCC19606; 750 μM for ATCC19606R) was pipetted onto a 5 mm diameter paper disk and allowed to air-dry. Prior to applying the drug-coated disk, 0.25 mL of bacterial suspension (OD₆₀₀ = 0.5) was spread onto an agar plate. The plates were incubated at 37 °C, and after 24 h, the diameter of the bacterial inhibition zone was measured using a vernier caliper. Ciprofloxacin (45 μM for ATCC19606; 750 μM for ATCC19606R) was used as a positive control, while normal saline was used as a negative control.

The MIC values were determined by the broth microdilution method. After determining the MIC, 0.25 mL of bacterial cultures with PMB concentrations higher than MIC were plated onto solid agar. Following incubation at 37 °C for 24 h, the number of colonies on the plates was observed, and the drug concentration that yielded fewer than 5 colonies was determined as the minimum bactericidal concentration (MBC) of PMB.

2.5. Detection of ferroptosis in A. baumannii induced by P/Fe@L-P

2.5.1. Affinity of P/Fe@L-P towards A. baumannii

To detect the affinity of P/Fe@L-P towards A. baumannii, a 1 mL of liposomes simultaneously labeled by FITC-PMB and RITC-cholesterol was added to 0.5 mL of bacterial suspension (OD₆₀₀ = 0.5) and incubated with shaking at 37 °C at 210 rpm. At various time intervals, the samples were centrifuged at 5,000 rpm for 10 min, and the affinity of the liposomes to bacteria was investigated by observing the fluorescence of the precipitate under ultraviolet (UV) light. Additionally, the centrifuged precipitate was observed with an inverted fluorescence microscope (IX73, Olympus, Tokyo, Japan) and a laser scanning confocal microscope (Leica STELLARIS 5, Wetzlar, Hessen, Germany).

2.5.2. GPX 4 and Nrf 2 detection

ATCC19606R (OD₆₀₀ = 0.5) was incubated with free PMB, P@L-P and P/Fe@L-P (0.015 mM). After 24 h, the bacteria were collected by centrifugation and lyzed by repeated freeze-thaw. The lysate was detected by Western blot assay. About 60 μg of the protein in each sample was subjected to 4%–12% polyacrylamide gel electrophoresis and subsequently transferred to the nitrocellulose membrane. After 1 h, the membrane was blocked, and then GPX4 and Nrf2 polyclonal antibodies were incubated with the membrane overnight at 4 °C, respectively. Subsequently, the membranes were incubated with horseradish peroxidase coupled secondary antibody for 1 h and then washed three times in tris buffered saline containing Tween 20. Image processing and analysis were performed using flour chemistry system (V37, Epson, Nagano-ken, Japan).

2.5.3. Glutathione (GSH) and malondialdehyde (MDA) measurements

After incubating the bacteria with the liposome for 24 h, bacterial precipitation was obtained by centrifugation (5,000 rpm, 10 min). Subsequently, 0.5 mL of normal saline was added, and the bacterial lysate was obtained by repeated freezing and thawing. The levels of GSH and MDA in bacteria were determined by GSH and MDA kit (Beyotime Technology, Shanghai, China).

2.5.4. ROS measurement

Bacteria (OD₆₀₀ = 0.5) was incubated with P/Fe@L-P (0.015 mM) and reagent from a ROS detection kit (10 μM; Beyotime Technology) at 37 °C with shaking (85 rpm). After 24 h, the qualitative observations of ROS levels in the bacteria were performed using an inverted fluorescence microscope, and quantitative analyses were carried out with a multi-functional microplate reader (BioTek Instruments, Inc., Winooski, VT, USA) with the excitation and emission wavelengths of 488 and 525 nm, respectively.

2.5.5. Fenton reaction detection

The methyl blue degradation test was used to determine the Fenton reaction activity of P/Fe@L-P. In brief, 100 μL of methyl blue solution (0.1 mg/mL) was added to a mixture of 1 mL of P/Fe@L-P and 1 mL of 0.01% hydrogen peroxide is prepared, and the UV–Vis absorption spectra of methyl blue at a wavelength of 644 nm were measured using a UV–Vis spectrophotometer (Lambda Bio40, PerkinElmer, Shanghai, China) at 2 h.

2.5.6. Hydroxyl free radical scavenging

Thiourea is a hydroxyl free radical blocking agent that can bind hydroxyl free radicals generated during the ROS generation [28]. The MIC and MBC of P/Fe@L-P in MHB medium containing thiourea (50 mM) were detected according to the methods described in Section 2.4, and the qualitative and quantitative detection of ROS were performed as described in Section 2.5.4.

2.5.7. Iron chelation

The iron chelator 2,2′-bipyridine was used to specifically bind Fe²⁺ which can inhibit ferroptosis by decreasing iron-dependent lipid peroxidation [29]. After pretreating bacteria with 2,2′-bipyridyl (50 μM) at 37 °C for 20 min, the MIC and MBC values of P/Fe@L-P were detected.

2.6. In vivo antibacterial efficacy study

2.6.1. Animals

Male Swiss mice (average weight 35 ± 5 g) were purchased from the Animal Experimental Center of Xuzhou Medical University (Xuzhou, China). The animal experiments were approved by the animal ethics committee of Xuzhou medical university (Approval No.: 202206S015). All animals were maintained under specific pathogen-free conditions.

2.6.2. Induction of mouse pneumonia model

First, the mice were injected intraperitoneally with cyclophosphamide (300 mg/kg). Four days later, 0.15 mL of ATCC19606R suspension (OD₆₀₀ = 0.5) was instilled into the lungs via a visualized tracheal cannula; this time point was designated as day 0. After 12 h, the mice exhibited symptoms of lethargy and tachypnea, indicating successful establishment of the model.

2.6.3. Drug administration

Fifty-four mice were randomly divided into 9 groups: (1) negative control (N–C); (2) pneumonia model (Model); (3) blank liposome via inhalation (Inh-B-L); (4) FeCl₂ solution via inhalation (Inh-Fe); (5) PMB solution via intravenous injection (IV-PMB); (6) PMB solution via inhalation (Inh-PMB); (7) P@L via inhalation (Inh-P@L); (8) P@L-P via inhalation (Inh-P@L-P); (9) P/Fe@L-P via inhalation (Inh-P/Fe@L-P). Twelve hours after bacterial infection, mice in each group were treated as follows: mice in the IV-PMB group received an intravenous injection of PMB (2 mg/kg) via the tail vein; mice in the Inh-B-L, Inh-PMB, Inh-P@L, Inh-P@L-P, and Inh-P/Fe@L-P groups were exposed to the certain liposome suspensions or PMB solution aerosols for inhalation. For the nebulization administration, a nebulizer for animals (YLS-8B, Yiyan Co., Jinan, China) was used; the nebulization condition was set as follows: six mice were put into an atomizing chamber (5 L); 10 mL of PMB solution or P/Fe@L-P suspension (1 mg/mL) was used for nebulization; the nebulization was carried out in a cycle mode of working for 1 min and resting for 4 min; the nebulization and inhalation were performed for 40 min, and this process could deliver about 0.04 mg PMB to the lung of the mouse, which had been detected by HPLC before. This treatment was administered once daily for 5 consecutive days; mice in the Inh-Fe group or the Model groups were nebulized with FeCl₂ solution (10 mL, 0.78 mg/mL) or normal saline using the same method. The difference between the N–C group and Inh-B-L group lies in the fact that mice in the N–C group were neither infected with bacteria nor subjected to any drug intervention, whereas mice in the Inh-B-L group were infected with bacteria and then inhaled B-L. The purpose of establishing the Inh-B-L group is to exclude the potential antibacterial effect of the blank liposomes.

2.6.4. Pharmacodynamic evaluation

From day 0 to day 7, the body weights of mice were measured daily, and the mortality rate of each group was recorded using a camera (Nikon, Tokyo, Japan) and the ANY-maze software. Blood oxygen saturation (SpO₂) was monitored using a small animal pulse oximeter (MouseOx, Shanghai Yuyan Instruments, Shanghai, China). Respiratory function was assessed using a small animal pulmonary function testing system (WBP-4R, TOW-INT Tech, Shanghai, China). The activity of mice was recorded every other day using a camera (ANY-maze, Shanghai, China).

On day 7, micro-computed tomography (micro-CT) (MadicLab Ultra, Linyi, China) was performed to observe changes in the lungs of the mice. Before the CT scan, iopamidol (930 mg/kg) was injected 15 min prior to improve imaging resolution. Following the CT scan, the blood was tested by a hematology analyzer; then, mice were euthanized, and a portion of the left lung from each mouse was collected for pathological examination after hematoxylin and eosin (H&E) staining. Additionally, interleukin-6 (IL-6) immunofluorescence detection in the left lung samples was performed using FITC-labeled anti-IL-6 antibodies (Elabscience Biotechnology Co. Ltd.).

The right lungs of mice were homogenized, and diluted samples were plated onto bacterial culture plates under aseptic conditions. Colony counts were performed to measure bacterial growth in the lungs.

2.7. Preliminary detection of biosafety

Eighteen male Swiss mice were randomly divided into 3 groups: (1) IV-PMB group; (2) PMB inhalation group; (3) P/Fe@L-P inhalation group. Mice in the IV-PMB group received daily intravenous injections of PMB (2 mg/kg) via the tail vein, while mice in the PMB inhalation and P/Fe@L-P inhalation groups received daily nebulized PMB or P/Fe@L-P, respectively, as described in Section 2.6.3. After 10 consecutive days of administration, blood was collected from the ocular globe of the mice, and serum creatinine and urea nitrogen levels were measured using an automatic biochemical analyzer (7600, Hitachi, Tokyo, Japan) to reflect renal toxicity. Following euthanasia, the hearts, livers, spleens, lungs, and kidneys were harvested, sectioned, and stained with H&E for pathological examination.

2.8. Statistical analysis

One-way analysis of variance (ANOVA) was used for comparisons among multiple groups. Depending on the homogeneity of variance, the least significant difference method or Dunnett's method was applied. The level of significance for hypothesis testing was set at α = 0.05, with differences considered statistically significant at P < 0.05 and highly significant at P < 0.01.

3. Results

3.1. Characterization of P/Fe@L-P

TEM observation (Fig. 2A) revealed that the synthesized P/Fe@L-P exhibited a spherical structure with good dispersion and the unique concentric structure of liposomes under TEM [30]. The particle size distribution of the liposomes was measured by dynamic light scattering. As shown in Fig. 2B, the average diameter of the liposome particles was 123.39 nm, with a polydispersity index of 0.303.

Fig. 2.

Characterization of P/Fe@L-P. (A) Transmission electron microscopy (TEM) images of P/Fe@L-P. (B, C) Particle size distribution (B) and zeta potential (C) as determined by dynamic light scattering. (D) The release of polymyxin B (PMB) from P/Fe@L-P in vitro. (E, F) The physical stability of P/Fe@L-P stored at 25 (E) or 40 °C (F). (G, H) Qualitative (G) and quantitative (H) observations of the ability of P/Fe@L-P to penetrate an artificial mucous layer. n = 3. P/Fe@L-P: PMB and Fe²⁺ loaded liposomes with PEGylation; EE (%): encapsulation efficiency (%).

The zeta potential of P/Fe@L-P was close to neutral (Fig. 2C). This was because the PMB was mainly encapsulated in the liposomes, and DPPC and cholesterol are electroneutral lipid components, resulting in the zeta potential of the prepared liposomes to be close to neutral. The similar result could also be observed in the previously reported literature [31].

Before the ultrafiltration, the EE% of PMB and Fe²⁺ in P/Fe@L-P were calculated to be 50% ± 3% and 55% ± 1%, respectively; after removing the free drug by ultrafiltration, the EE% of PMB and Fe²⁺ in the purified system reached 99%, proving the free drug had been removed from the system. The LE% were 10.8% ± 0.9% and 8.4% ± 0.5%, respectively. The effects of the formulation of P/Fe@L-P on its particle size and EE% are shown in Figs. S1–S3.

As depicted in Fig. 2D, the release rate of PMB from P/Fe@L-P was rapid within the first 12 h, resulting in a cumulative release of 80% of PMB within this period. From 12 to 48 h, the release rate slowed down, and ultimately, approximately 95% of PMB was released from P/Fe@L-P within 48 h.

To investigate the physical stability of the particles, particle size and EE% were measured during incubation at 25 °C or 40 °C for 30 days. After 30 days of storage at both temperatures (Figs. 2E and F), the particle size of the liposomes remained steady, potentially reflecting no aggregation appeared. What's more, the EE% of the liposomes remained unchanged, further indicating the high physical stability of P/Fe@L-P.

Fig. 2G illustrates the penetration of P/Fe@L-P liposomes with different diameters through an artificial mucus layer. Liposomes with diameters of 150 and 250 nm traversed the mucus layer faster than those with a diameter of 500 nm, suggesting that reduced particle size enhances permeability. The restricted movement of particles larger than 500 nm is consistent with the known mesh size of the mucus layer (approximately 100–500 nm) [32,33]. Quantitative measurements (Fig. 2H) also demonstrated that 150 nm liposomes diffused through the mucus layer faster than those with diameters of 250 and 500 nm.

3.2. Antibacterial effect of P/Fe@L-P against A. baumannii in vitro

3.2.1. Antibacterial effect of P/Fe@L-P against ATCC19606

The antibacterial effect of P/Fe@L-P was determined through the inhibition zone assay on solid medium. As shown in Figs. S4A and B, when applied to the wild-type strain, the diameter of the inhibition zone produced by free PMB (12.4 ± 0.2 mm) was larger than that produced by liposomal PMB formulations (P@L: 11.0 ± 0.3 mm, P@L-P: 11.0 ± 0.1 mm, P/Fe@L-P: 11.0 ± 0.2 mm). As shown in Figs. S4C and D, the MIC and MBC values of PMB against ATCC19606 were 1.5 and 3.0 μM, respectively, which were lower than those achieved with P/Fe@L-P treatments (3.0 and 6.0 μM, respectively). In summary, these results indicate that free PMB exhibits stronger antibacterial effects against the wild-type A. baumannii strain than the encapsulated drug. This difference may be attributed to the ability of free PMB to directly bind to LPS on the bacterial surface, thereby functioning more effectively than encapsulated PMB.

3.2.2. Antibacterial effect of P/Fe@L-P against ATCC19606R

Fig. S5 shows that the resistance to PMB was induced in A. baumannii, and this strain was named ATCC19606R. In details: after gradual increases in PMB concentration during co-incubation with bacteria, the MIC increased from 1.2 to 72 μM (Fig. S5A); the cell wall thickness (including the lipopolysaccharide layer, peptidoglycan layer, and plasma membrane) decreased to 33 ± 1 nm from 50 ± 1 nm, and the LPS layer decreased to 17 ± 1 nm from 25 ± 2 nm (Fig. S5B); the zeta potential of the bacteria was also decreased, consistent with the reduction in LPS thickness.

As shown in Figs. 3A and B, when testing ATCC19606R, the inhibition zone diameter of P@L was 13.0 ± 0.05 mm, larger than that of free PMB (11.5 ± 0.02 mm); the MIC and MBC values of P@L were 60 μM and 384 μM (Figs. 3C and D), respectively, both lower than those of free PMB (MIC: 72 μM, MBC: 768 μM); these results indicated that liposomal encapsulation enhanced the inhibitory capability of PMB against resistant strains.

Fig. 3.

P/Fe@L-P possesses a more effective antibacterial action against polymyxin B (PMB)-resistant A. baumannii (ATCC19606R) as compared with free PMB and other PMB-loaded liposomes. (A, B) The inhibition zones were detected qualitatively (A) and quantitatively (B) with ciprofloxacin (750 μM) as the positive control and normal saline as the negative control. (C–E) Determinations of minimum inhibitory concentration (MIC) (C), minimum bactericidal concentration (MBC) (D) and growth curve (E) of P/Fe@L-P. (F) Transmission electron microscopy (TEM) images of ATCC19606R after treatment with P/Fe@L-P (2.7 μM) for 12 h. N–C: negative control; OD₆₀₀: optical density at 600 nm; P@L: PMB loaded liposomes; P@L-P: PMB loaded liposomes with PEGylation; P/Fe@L-P: PMB and Fe²⁺ loaded liposomes with PEGylation. ∗∗P < 0.01; ∗∗∗P < 0.001. n = 3.

The MIC and MBC values of P@L-P were similar to those of P@L as shown in Figs. 3C and D, indicating that surface PEGylation had no detectable effect on the antibacterial ability of PMB liposomes in vitro. Compared to P@L and P@L-P, P/Fe@L-P exhibited the largest inhibition zone diameter and the lowest MIC and MBC values (54 and 192 μM, respectively), indicating that co-encapsulated Fe²⁺ in liposomes potentiates the antibacterial effect of PMB.

In this study, we observed that the MIC of P/Fe@L-P against ATCC19606R was 54 μM, which is significantly higher than the clinical breakpoint for PMB (2 μg/mL, equivalent to 1.5 μM). This finding may be attributed to the influence of bacterial resistance on MIC values. In this research, we induced A. baumannii to PMB-resistant strain and increased its MIC value from 1.2 to 72 μM. To rigorously assess the antibacterial efficacy of P/Fe@L-P, we selected strains with the highest MIC values for our experiments. Our results revealed that when the MIC values of free PMB were 9 and 18 μM, the MIC of P/Fe@P-L was 1.5 μM, which is equivalent to the clinical breakpoint value. For strains with higher MIC values of free PMB at 36 and 72 μM, although the MIC of P/Fe@P-L (6 and 54 μM) exceeded the clinical breakpoint, it was still notably lower than that of free PMB. This indicates that P/Fe@P-L possesses superior bactericidal ability compared to free PMB. Notably, a similar phenomenon of MIC values exceeding the clinical breakpoint has been reported for marketed amikacin liposomes, with a MIC value of 9 μg/mL against S. aureus [34], exceeding its breakpoint value of 8 μg/mL. Collectively, these findings strongly suggest the potential clinical value of P/Fe@P-L.

Analysis of bacterial growth curves cultivated in liquid medium in the presence of various drugs (Fig. 3E) revealed that the lag phase of bacterial growth (0–8 h) was shorter for bacteria treated with free PMB (2.7 μM) than for those treated with P/Fe@L-P with the same concentration (0–12 h). Similarly, the number of bacteria in the logarithmic growth phase (8–20 h) and stationary phase (20–24 h) was higher for bacteria treated with free PMB compared to those treated with P/Fe@L-P (12–20 h and 20–24 h, respectively). These results further support the stronger inhibitory effect of P/Fe@L-P against PMB-resistant A. baumannii compared to free PMB.

As shown in Fig. 3F, after treated by P/Fe@L-P (2.7 μM) for 12 h, the outer edge of the bacterial wall became more reticulated, the inner membrane became thinner, and intracellular components such as DNA, lipids, and proteins were disrupted, leading to bacterial lysis and death.

3.3. Antibacterial effect of P/Fe@L-P against multidrug-resistant K. pneumoniae and S. aureus

In order to verify the efficient bactericidal effect of P/Fe@L-P against a variety of drug-resistant bacteria, we selected another Gram-negative bacterium such as K. pneumoniae and a Gram-positive bacterium such as S. aureus to carry out more extensive bactericidal tests.

Fig. 4A displays the growth of multidrug-resistant K. pneumoniae in liquid medium after treatment with PMB or P/Fe@L-P. The results indicate that the MIC value of P/Fe@L-P is 3 μM, lower than that of free PMB (6 μM). In the growth curve of K. pneumoniae treated with 0.75 μM P/Fe@L-P (Fig. 4B), the lag phase lasts from 0 to 9 h post-inoculation, followed by the logarithmic phase from 9 to 18 h, and the stationary phase from 18 to 24 h. In contrast, that treated with an equivalent amount of P/Fe@L-P shows delayed phases: a lag phase (0–12 h), logarithmic phase (12–20 h), and stationary phase (20–24 h). In this study, given that bacterial growth is fully inhibited at the MIC concentration, a sub-MIC concentration was utilized to assess the growth curve. As shown in Fig. 4C, the MBC value of P/Fe@L-P against K. pneumoniae is 24 μM, lower than that of PMB (48 μM). These results demonstrate that P/Fe@L-P exhibits stronger inhibitory effects against multidrug-resistant K. pneumoniae than free PMB.

Fig. 4.

P/Fe@L-P exhibits bactericidal activity against multidrug-resistant K. pneumoniae not against multidrug-resistant S. aureus. (A, B) Growth curves of multidrug-resistant K. pneumoniae incubated with P/Fe@L-P of various concentrations for 18 h (A) and with P/Fe@L-P (0.75 μM) for 24 h (B). (C) Minimum bactericidal concentration (MBC) of P/Fe@L-P against K. pneumoniae. (D, E) Growth curves of multidrug-resistant S. aureus treated with P/Fe@L-P of various concentration for 18 h (D) and with P/Fe@L-P (3 μM) within 24 h (E). (F) MBC of P/Fe@L-P against S. aureus. OD₆₀₀: optical density at 600 nm; PMB: polymyxin B; P/Fe@L-P: PMB and Fe²⁺ loaded liposomes with PEGylation.

The bactericidal test in S. aureus is shown in Figs. 4D–F. The results show that both P/Fe@L-P and PMB have the same MIC (12 μM) and MBC (96 μM) values against multidrug-resistant S. aureus. The growth phases of bacteria treated with PMB and P/Fe@L-P (3 μM, Fig. 4E) are also identical, with a lag phase from 0 to 4 h post-inoculation, a logarithmic phase from 4 to 18 h, and a stationary phase from 18 to 24 h. Compared with Gram-negative bacteria, the poor killing effect of P/Fe@L-P against S. aureus may be attributed to the dense peptidoglycan layer on the surface of S. aureus, which is expected to form a physical barrier inhibiting the entry of P/Fe@L-P, thereby reducing its bactericidal effect.

These results indicate that P/Fe@L-P is effective in killing Gram-negative bacteria such as polymyxin-resistant A. baumannii and multidrug-resistant K. pneumoniae. However, due to the presence of a dense peptidoglycan layer, P/Fe@L-P does not exhibit significant antibacterial effects against Gram-positive bacteria like S. aureus. Therefore, developing strategies to penetrate the peptidoglycan layer and eradicate Gram-positive bacteria is crucial for further optimizing this system.

3.4. Induction of ferroptosis by P/Fe@L-P

3.4.1. Internalization of P/Fe@L-P in ATCC19606R

To investigate the interaction between PMB and bacteria, FITC-labeled free PMB or various liposomal formulations were incubated with bacteria for 2 h, followed by bacterial collection and washing via centrifugation. As shown in Fig. 5A, when the ATCC19606 strain was incubated with liposomal FITC-PMB, weaker fluorescence was observed in the precipitate compared to bacteria incubated with free FITC-PMB, showing the more efficient binding of free PMB to ATCC19606 than liposomal formulations. However, for the resistant strain (ATCC19606R), bacteria treated with FITC-PMB in liposomal formulations exhibited stronger fluorescence signals than those treated with free FITC-PMB, indicating that liposomes facilitated the binding or internalization of PMB in ATCC19606R. These results are consistent with the antibacterial activities shown in Fig. 3.

Fig. 5.

Encapsulation in liposomes enhanced the internalization of polymyxin B (PMB) by PMB-resistant A. baumannii (ATCC19606R). (A) The binding and internalization of P/Fe@L-P carrying fluorescein isothiocyanate (FITC)-PMB by ATCC19606 and ATCC19606R as detected using an ultraviolet lamp. (B) The affinity of rhodamine B isothiocyanate (RITC)-labeled liposomes containing FITC-PMB for ATCC19606 and ATCC19606R observed by laser confocal microscopy (the bacteria was labeled by FITC). (C) The colocation of FITC-PMB with RITC-labeled liposomes upon incubation with ATCC19606 and ATCC19606R as observed by inverted fluorescence microscopy. (D) The internalization of P/Fe@L-P in ATCC19606R observed by transmission electron microscopy (TEM); false color was added using Photoshop. P@L: PMB loaded liposomes; P@L-P: PMB loaded liposomes with PEGylation; P/Fe@L-P: PMB and Fe²⁺ loaded liposomes with PEGylation.

Further analysis of the interaction between liposomes themselves and bacteria was conducted by labeling the liposomes with RITC-cholesterol and labeling bacteria by FITC. In contrast to the pronounced binding of PMB to ATCC19606, little RITC-labeled liposome fluorescence was observed in ATCC19606 (Fig. 5B). Conversely, when ATCC19606R was incubated with these labeled liposomes, stronger red fluorescence was detected, suggesting the high affinities of liposomes towards PMB-resistant bacteria.

To further elucidate the role of liposomes in the binding and entry of PMB into bacteria, bacteria were treated with liposomes simultaneously labeled with both the FITC-PMB and the RITC-cholesterol. As shown in Fig. S6Figs. 5C and S6, consistent with the data in Fig. 5A, free PMB exhibited the strongest binding ability with wild-type bacteria; compared to the wild-type strain, less free PMB was found to bind to PMB-resistant bacteria; this difference in binding may be attributed to the reduced LPS in the membrane of ATCC19606R. Compared to free PMB, P@L, P@L-P, and P/Fe@L-P resulted in more pronounced FITC-PMB fluorescence signals in the ATCC19606R strain, and most of the fluorescence signals from liposomes overlapped with the green fluorescence of PMB in bacteria, with only a small portion of PMB fluorescence signals present alone, suggesting that most of the PMB interacting with these cells was embedded in liposomes, but a small fraction of interacting PMB had been released from the liposomes.

Similarly, TEM analysis revealed the presence of liposomes within ATCC19606R after incubation (Fig. 5D) as compared with the negative control strain, further corroborating the affinity of P/Fe@L-P towards PMB-resistant A. baumannii.

3.4.2. GPX4 and Nrf2 expressions in ATCC19606R

As depicted in Figs. 6A and B, upon interaction with ATCC19606R, P@L, P@L-P, and P/Fe@L-P significantly downregulated the expression of bacterial GPX4 and Nrf2 proteins, indicating that all three formulations triggered ferroptosis in bacteria. Notably, compared to P@L, the expressions of Nrf2 and GPX4 in bacteria treated with P@L-P were markedly reduced, likely due to the enhanced penetration of PMB into bacterial cells facilitated by liposomes (Fig. 5). Furthermore, the changes in Nrf2 and GPX4 expressions were more pronounced in the P/Fe@L-P group compared to P@L-P, suggesting that the co-delivery of Fe²⁺ significantly potentiated the ferroptosis-inducing effect of PMB.

Fig. 6.

P/Fe@L-P induced ferroptosis in ATCC19606R due to enhanced catalysis by Fe²⁺. (A, B) Qualitative (A) and quantitative (B) detections of GPX4 and Nrf2 expressions in ATCC19606R after incubating with P/Fe@L-P (15 μM) for 24 h. (C, D) Glutathione (GSH) (C) and malondialdehyde (MDA) (D) levels in ATCC19606R after incubating with P/Fe@L-P (15 μM) for 24 h. (E, F) Qualitative (E) and quantitative (F) detections of ROS generation in ATCC19606R. (G–J) Hydroxyl radical scavenging by thiourea (T) inhibited the reactive oxygen species (ROS) generation in ATCC19606R (G, H), and increased the minimum inhibitory concentration (MIC) (I) and minimum bactericidal concentration (MBC) (J) values of P/Fe@L-P. (K–N) Iron scavenging by 2,2′-bipyridine (2,2′-B) inhibited the ROS generation in ATCC19606R (K, L) and increased the MIC (M) and MBC (N) values of P/Fe@L-P. PMB: polymyxin B; B-L: blank liposome; B-L + P: physical mixture of blank liposome and free PMB; P@L: PMB loaded liposomes; P@L-P: PMB loaded liposomes with PEGylation; P/Fe@L-P: PMB and Fe²⁺ loaded liposomes with PEGylation; GAPDH: glyceraldehyde-3-phosphate dehydrogenase. ^∗P < 0.05; ^∗∗P < 0.01; ^∗∗∗P < 0.001. ns: no significance. n = 3.

In this study, due to the unavailability of antibodies specifically targeted at GPX4, Nrf2, and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) in prokaryotic cells, we opted to utilize antibodies designed for eukaryotic cells. Fortunately, the expression of these three proteins was successfully detected, which also confirmed their presence in A. baumannii. This finding aligns with previous literature [35,36] reports that GPX-4 and NADPH can indeed be detected in bacterial species.

3.4.3. GSH and MDA levels in ATCC19606R

Quantitative assessments of bacterial GSH and MDA expressions were performed. The results (Figs. 6C and D) revealed that both P@L-P and P/Fe@L-P significantly decreased the intracellular GSH levels in bacteria compared to the saline group. Notably, P/Fe@L-P exhibited a stronger capacity to reduce GSH levels than P@L-P, while simultaneously demonstrating the most potent ability to elevate intracellular MDA levels. These findings align with the changes in GPX4 and Nrf2 expressions, further corroborating the effective induction of ferroptosis in ATCC19606R by the co-delivery of PMB and Fe²⁺.

3.4.4. ROS generation in ATCC19606R

Fluorescent probes were employed to qualitatively and quantitatively detect ROS generated by PMB or liposomal PMB formulations. In the drug-resistant ATCC19606R strain (Figs. 6E and F), free PMB generated lower ROS levels than in wild-type bacteria (Fig. S7), which aligns with its weaker binding and internalization capabilities in resistant strains (Fig. 5). Treatment with liposomal P@L, P@L-P, and P/Fe@L-P led to increased ROS production in ATCC19606R, with the most pronounced effect observed in the P/Fe@L-P group, further confirming the most effective ferroptosis induction by P/Fe@L-P observed in Figs. 6A and B.

To demonstrate that the generation of ROS was induced by the Fenton reaction catalyzed by P/Fe@L-P, the methyl blue degradation test was performed. As illustrated in Fig. S8, both the FeCl₂ group and the P/Fe@L-P group exhibited significant degradation of methyl blue, providing evidence for the occurrence of the Fenton reaction.

In this study, a weaker yet obvious ROS signal was observed in the P@L-P group, which confirmed the ability of liposomal PMB to generate ROS through the Fenton reaction within cellular environments. Conversely, the P/Fe@L-P group exhibited a marked enhancement in ROS production, attributed to the catalysis of the Fenton reaction by Fe²⁺.

3.4.5. Hydroxyl free radical scavenging

To further substantiate the potential role of ferroptosis induction in the antibacterial activity of P/Fe@L-P, thiourea was utilized as a hydroxyl radical scavenger to inhibit ferroptosis. As shown in Figs. 6G and H, ROS generation in all treatment groups was completely inhibited by co-treatment with thiourea. Under these conditions, where intracellular ROS production and subsequent ferroptosis were suppressed, the MIC (Fig. 6I) and MBC (Fig. 6J) values of liposomal formulations significantly increased. These results demonstrate that P/Fe@L-P exerts its bactericidal effect through inducing ferroptosis rather than the traditional mechanism of binding to bacterial surface LPS.

3.4.6. Iron chelation

To further validate the role of Fe²⁺ co-delivery in P/Fe@L-P-induced bacterial ferroptosis, 2,2′-bipyridine was used as an iron chelator. As illustrated in Figs. 6K and L, when co-treated with 2,2′-bipyridine, ROS production by P@L and P/Fe@L-P in ATCC19606R decreased. Correspondingly, the antibacterial efficacy of P@L and P/Fe@L-P against ATCC19606R was reduced, as evidenced by increased MIC and MBC values (Figs. 6M and N). These findings indicate that Fe²⁺ co-delivery enhances the Fenton reaction and ROS generation in PMB-resistant A. baumannii, thereby potentiating the ferroptosis-inducing effect of P/Fe@L-P (Fig. 6A), leading to the enhanced antibacterial activity of liposomes observed in Fig. 3.

3.5. Treatment of ATCC19606R-induced pneumonia in mice

3.5.1. Therapeutic effects on physical manifestations of pneumonia mice

A mouse model of pneumonia was induced by administering ATCC19606R into the lungs as shown in Fig. 7A. The mental status (Fig. 7A) and the moving trajectories (Fig. 7B) of mice during the treatment process show that the model group mice exhibited disheveled fur and lethargy after bacteria infusion, and signs of infection improved in mice in the Inh-PMB, Inh-P@L, Inh-P@L-P, and Inh-P/Fe@L groups obviously.

Fig. 7.

Inhalation of P/Fe@L-P protected the pneumonia model mice infected with ATCC19606R. (A–D) the lethargic mental state (A), movement (B), survival rates 704 (C), and CT imaging properties (D) of the mice. n = 6. N–C: negative control; Iv: intravenous injection; Inh: inhalation; PMB: polymyxin B; P@L: PMB loaded liposomes; P@L-P: PMB loaded liposomes with PEGylation; P/Fe@L-P: PMB and Fe²⁺ loaded liposomes with PEGylation.

The survival curves of these mice are presented in Fig. 7C. Mice in the Inh-Fe, Inh-B-L, and Model groups all died within 3 day, demonstrating the rapid progression of ATCC19606R-induced pneumonia. On day 7, the survival rate of mice in the Inh-PMB group was 50%, higher than that of the Iv-PMB group (33%). This comparison underscores the superiority of pulmonary inhalation over intravenous injection and provides a rationale for the clinical application of inhaled PMB. However, it is worth noting that compared to the Inh-PMB group, inhalation of P@L and P@L-P (Inh-P@L and Inh-P@L-P groups) led to a delayed mortality rate in mice. This delay could be attributed to the enhanced antibacterial effect of liposomal PMB, as illustrated in Fig. 3, as well as the prolonged retention of the drug in the liposomal formulations, as evidenced in Fig. S9. Despite this, the survival rates within the first 7 day for these two liposomal groups did not surpass those of the non-liposomal formulations.

Compared to the Inh-P@L group, treatment with PEGylated P@L-P liposomes (Inh-P@L-P group) resulted in delayed mouse mortality, potentially due to PEGylation enhancing liposome permeability through the mucus layer (Fig. 2H). Inhalation of P/Fe@L-P (Inh-P/Fe@L-P group) led to the highest survival rate (83.3%) on day 7, indicating this intact liposome was the most effective method for treating ATCC19606R-induced pneumonia. These results align with the superior antibacterial effect of P/Fe@L-P against ATCC19606R (Fig. 3) and its prolonged lung retention time (Fig. S9).

As expected, the body weights of mice in all groups significantly declined 2 day post-lung infection (Fig. S10). After 2–3 day of intravenous PMB administration or inhalation of PMB, P@L, P@L-P, or P/Fe@L-P, body weights began to gradually increase in these groups, and among all treatments, P/Fe@L-P had the greatest impact on body weight recovery in mice, further supporting its effectiveness in treating ATCC19606R-induced pneumonia.

The results of blood test (Table S3) showed that compared to normal mice, the C-reactive protein levels in the Model group mice were significantly elevated. However, following treatment with P/Fe@L-P, the C-reactive protein levels in the mice returned to normal, demonstrating the therapeutic effect of P/Fe@L-P on bacterial infectious pneumonia. Since the Model group mice were treated with cyclophosphamide, their lymphocyte counts were significantly lower compared to the normal group. Even after treatment, the lymphocyte counts remained low, which may be attributed to the damage caused by cyclophosphamide to the immune system.

3.5.2. Lung imaging

12 h post-infection with ATCC19606R, and prior to any pharmacological intervention, CT images of the mice lungs (Fig. 7D) revealed notable differences between the Model group and the N–C group. Specifically, the Model group exhibited pronounced ground-glass opacity (highlighted by blue circles) and indistinct lung tissue contours (indicated by red arrows), indicative of bacterial infection-induced lung injury. It is worth noting that, at this juncture, the mice in the Model group mirrored the condition of the other groups, excluding the N–C group; thus, only the CT findings of the Model group are presented here.

In the Iv-PMB and Inh-PMB groups, abnormal lung morphology could still be observed until day 7. In the Inh-P@L-P group, the coarsened lung markings were alleviated by day 4, with clearer lung contours; however, notable ground-glass opacity and thickened bronchial walls (highlighted by yellow arrows) remained; by day 7, the ground-glass opacity disappeared, but bronchial wall thickening was still observable. Compared to the Inh-P@L-P group, mice in the Inh-P/Fe@L-P group showed clearer lung markings on day 4; nonetheless, minor ground-glass opacity and bronchial wall thickening persisted; by day 7, the ground-glass opacity disappeared, and lung imaging features resembled those of normal mice, indicating that inhalation of P/Fe@L-P reversed the lung structural changes induced by ATCC19606R infection more effectively than other samples. Fig. S11 further proved P/Fe@L-P improved the respiratory function of the pneumonia mice effectively.

Due to the mortality of mice in the Model, Inh-Fe, and Inh-B-L groups on day 3, only the results from day 2 are presented. On this day, CT scans revealed extensive areas of ground-glass opacity, coarse lung markings, thickened bronchial walls, and indistinct lung tissue contours. These observations indicate that the bacterial infection was not effectively inhibited in these groups.

3.5.3. Lung bacterial load and pathological changes

The lung bacterial load in mice post-treatment was quantified as an indicator of antibacterial activity (Figs. 8A and B). Results indicated that significant bacterial populations persisted in the lungs of model mice and those treated with Inh-Fe and Inh-B-L. In contrast, reduced lung bacterial levels were observed in mice treated with Iv-PMB, Inh-PMB, Inh-P@L, Inh-P@L-P, and Inh-P/Fe@L-P, particularly in the Inh-P/Fe@L-P group, where no viable bacteria were detected in the lungs, suggesting the most potent bactericidal effect of P/Fe@L-P in vivo. This effect directly correlated with the most significant improvements in survival rate, lung tissue structure, respiratory function, and lung inflammation by P/Fe@L-P.

Fig. 8.

Inhalation of P/Fe@L-P completely killed ATCC19606R infecting the lung and reversed inflammation and pathological injury of the lung in the pneumonia mice. (A, B) Bacterial numbers isolated from lungs of infected mice. (C) Immunofluorescent detection of interleukin-6 (IL-6) expression in the lung of mice. (D) Pathological examination of lung tissues following hematoxylin and eosin (H&E) staining. Abbreviation: B-L: blank liposome; N–C: negative control; iv: intravenous injection; inh: inhalation; PMB: polymyxin B; P@L: PMB loaded liposomes; P@L-P: PMB loaded liposomes with PEGylation; P/Fe@L-P: PMB and Fe²⁺ loaded liposomes with PEGylation. The symbol "#" means that the total bacterial count is higher than 5 × 10³.

According to immunofluorescence analysis (Fig. 8C), a substantial amount of pro-inflammatory cytokine (IL-6) was present in the lungs of model mice post-ATCC19606R infection. Intravenous PMB administration slightly alleviated inflammation, while inhalation of P@L, P@L-P, and P/Fe@L-P reduced pro-inflammatory cytokine expression more effectively than other treatments. These findings concurred with the survival status (Fig. 7) and respiratory function recovery (Fig. S11) of mice.

Pathological examination (Fig. 8D) revealed alveolar septal hyperplasia leading to the disappearance of alveolar spaces (marked with black boxes), thickened airway walls (marked with black arrows), and massive inflammatory cell infiltration (marked with yellow boxes) in the lungs of pneumonia model mice. Treatment with Inh-Fe and Inh-B-L did not reverse these pathological changes. Iv-PMB partially reversed alveolar loss, but pulmonary interstitial thickening, ductal wall thickening, and inflammatory cell infiltration persisted. Compared to the Iv-PMB group, Inh-PMB treatment preserved more alveolar structures and mitigated inflammatory infiltration. Compared to the Inh-PMB group, more alveolar structures were observed in the Inh-P@L and Inh-P@L-P groups. Among all groups, mice in the Inh-P/Fe@L-P group exhibited the most normal pathological features, including thin alveolar walls, abundant normal alveolar structures, and thin bronchial walls, indicating complete reversal of lung damages.

3.6. Biocompatibility of P/Fe@L-P

The cytotoxicity of P/Fe@L-P was detected in Calu-3 cells, which were usually as a pulmonary epithelial model [37]. The results (Fig. S12) proved that the liposomal encapsulation decreased the cytotoxicity of free PMB significantly. The hemolysis rate of P/Fe@L-P within the concentration of 1.2–6.2 mg/mL was lower than 1% (Fig. S13), further proving the biocompatibility of P/Fe@L-P.

After 10 consecutive days of drug administration, mice were euthanized, and their blood and tissues were collected to investigate the biosafety of P/Fe@L-P treatment. As shown in Fig. S14A, compared with the negative control group, intravenous injection of free PMB for 10 day led to a significant increase in serum creatinine and urea nitrogen. Moreover, pathological sections stained with H&E (Fig. S14B) revealed swollen tubular epithelial cells in the PMB group, resulting in a markedly reduced lumen of renal tubules, along with evident proteinaceous fluid exudation observed within the tubules (marked by black arrows). These findings collectively indicate the nephrotoxicity of intravenous PMB. In contrast to the Iv-PMB group, the Inh-PMB and Inh-P/Fe@L-P groups exhibited decreased levels of serum creatinine and urea nitrogen, with clear and intact tubular structures observable in pathological sections. This may be attributed to the inhalation route enabling more drugs to reach the lungs of mice, thereby reducing drug distribution in the kidneys (Fig. S9), as well as shielding effect of liposome on positive charge of PMB as proved by the Zeta potential detections (Fig. 2C).

Lung pathological changes were also observed after 10 consecutive days of treatment. Results showed clear single-layered alveolar structures in the lungs of mice in the negative control group. However, in the Iv-PMB group, alveolar walls were thickened, with evident alveolar hemorrhage (marked with blue arrows) and widespread inflammatory cell infiltration (marked with yellow arrows), indicating that long-term PMB administration caused lung injury. In mice treated with Inh-PMB and Inh-P/Fe@L-P, the normal alveolar structures were observed; this reduced lung damage might be attributed to liposome encapsulation and the decreased direct interaction between PMB and the pulmonary epithelial cells.

Upon performing hepatic histopathologic analyses, it was observed that compared with the tight arrangement of hepatic cells in the negative control group, the cytoplasmic staining of the liver stem cells in the Iv-PMB group was lighter and the volume of the hepatic cells increased, suggesting the occurrence of hepatocyte edema. In tissues from mice treated with Inh-PMB or Inh-P/Fe@L-P, normal hepatic cells were observed, showing that the encapsulation of liposomes decreased the liver toxicity of PMB.

The pathological examination of spleen tissues showed that, as compared with the large number of tightly distributed lymphocytes in the negative control group, a large number of macrophages (marked by black hollow arrows) appeared in the Iv-PMB group, indicating that long-term intravenous injection of PMB activated the immune response. As compared with the Iv-PMB group, fewer macrophages could be observed in the spleen tissues of the Inh-PMB and Inh-P/Fe@L-P groups, indicating that liposome encapsulation relieved the immune stimulation induced by PMB.

Cardio pathological examination revealed no obvious damage to heart tissues in any treatment group.

4. Discussion

The overuse of antibiotics has led to a crisis of resistance, creating a significant challenge for global public health. In response, the scientific community has focused on developing new antibiotics with novel chemical structures. However, this effort is often hindered by high costs, long timelines [38], and the risk of triggering further resistance. To address this, our study presents a new strategy using PMB, through a drug delivery system that redefines the interaction between drugs and pathogens. Our goal is to restore the bactericidal effectiveness of older antibiotics and improve resource utilization.

Specifically, we have successfully used liposomes as carriers to deliver PMB and Fe²⁺. This approach shifts PMB's target from the bacterial surface LPS to the bacterial interior, changing its mechanism of action. It also promotes a synergistic effect between PMB and Fe²⁺ inside the bacterial cells, triggering ferroptosis, a lethal cellular demise pathway. This strategy not only enhances PMB's effectiveness against the PMB-resistant A. baumannii, but also offers a more cost-effective and faster path to approval via the 505(b)(2) pathway [39], avoiding the lengthy new drug development processes typically required by the FDA or national drug regulatory agencies. This accelerates time-to-market and could help meet the urgent need for new antibiotic research, providing a practical solution to the global antibiotic resistance crisis.

In the wild strain, it was observed that the free PMB could bind with the bacteria effectively; however, the bacteria inhibited the bind of the liposome. This was perhaps because that the expression of LPS on the surface of the wild strain LPS establishes a formidable physical barrier that hinders liposomes from penetrating through the bacterial wall [40], effectively blocking the ingress of these electrically neutral liposomes into the bacterial interior. Conversely, in strains resistant to PMB, the reduced expression of LPS leads to a marked enhancement in the permeability of liposomes towards bacteria.

In the bactericidal experiment, we show that P/Fe@L-P is more effective than free PMB against PMB-resistant A. baumannii (Fig. 3). In a mouse model mirroring clinical conditions, direct pulmonary inhalation of P/Fe@L-P led to significantly better outcomes in treating pneumonia caused by drug-resistant A. baumannii, compared to traditional intravenous administration of free PMB. This improved performance is due to three key factors: firstly, resistant strains reduce the affinity of free PMB for bacterial membranes by downregulating LPS expression (Fig. S5), decreasing its antibacterial potency; second, liposome encapsulation allows more efficient delivery of PMB and Fe²⁺ into bacterial cells (Fig. 5); and third, the Fenton reaction catalyzed by Fe²⁺ increases ROS production within bacteria, triggering ferroptosis and eliminating pathogens (Fig. 6).

It is notable that for non-resistant wild-type A. baumannii, liposomal encapsulation reduced the in vitro antibacterial activity of PMB (Fig. S4), likely due to the ability of free PMB to directly target LPS on the bacterial surface (Fig. 5). However, in the pneumonia treatment paradigm, P/Fe@L-P exhibited therapeutic efficacy comparable to intravenous free PMB against wild-type strains Fig. S15Fig. S17(Figs. S15–S17). This might be because that the wild-type strains are relatively mild, making it hard to distinguish between treatment methods in less severe infections. This finding highlights the broad-spectrum potential of P/Fe@L-P in managing A. baumannii-induced pneumonia, regardless of drug resistance.

To explore the broad-spectrum antibacterial effects of P/Fe@L-P, we tested its activity against a range of multidrug-resistant bacterial strains. The results showed that this liposomal system was highly effective against gram-negative bacteria, particularly K. pneumoniae (Fig. 4). However, its effect on Gram-positive bacteria, especially S. aureus, was less pronounced, likely due to the dense peptidoglycan layer on their outer surfaces, which restricts liposome penetration. Therefore, optimizing liposome formulations to improve their ability to cross the peptidoglycan barrier is an important direction for future research, aiming to broaden the antibacterial spectrum of P/Fe@L-P.

Clinically, the use of polymyxin B has been limited by its nephrotoxic side effects, with incidence rates reaching up to 60% in treated patients, posing a significant challenge for effective treatment protocols [41]. This study demonstrates that P/Fe@L-P, when administered via pulmonary inhalation, significantly reduce the nephrotoxic effects compared to traditional intravenous delivery of free PMB (Fig. S14). This benefit arises from two factors: first, pulmonary inhalation limits the distribution of PMB in the kidneys (Fig. S9), protecting them from direct exposure; second, liposome encapsulation reduces the drug's cytotoxicity, further protecting kidney function. Additionally, P/Fe@L-P reduced the risk of lung injury compared to inhalation of free drugs (Fig. S14), as encapsulation prevents PMB from interacting directly with lung epithelial cell membranes, avoiding potential damage. This improvement in biocompatibility provides a solid foundation for the use of high-dose, long-term PMB treatment in clinical settings.

Ferroptosis, a novel iron-dependent programmed cell death pathway from apoptosis and autophagy, has attracted significant attention in the biomedical community. Our study showed an increase in ROS levels in A. baumannii after treatment with P/Fe@L-P, along with a decrease in Nrf2 and GPX4 expression, increased MDA levels, and reduced GSH levels in the bacteria. These changes suggest the ferroptosis pathway was activated. To confirm this mechanism, we conducted iron chelation and ROS scavenging experiments, which revealed a significant reduction in the bactericidal effect of P/Fe@L-P under these conditions, confirming P/Fe@L-P's ability to induce ferroptosis in bacteria. Notably, while ferroptosis has been thoroughly explored in various eukaryotes since its inception by Dixon, et al. in 2012 [42], its presence in S. aureus and Escherichia coli has been reported by relatively fewer literatures [35,43,44], and this study proved in A. baumannii the ferroptosis could be induced by external manipulation, providing a new strategy for combating this pathogenic microorganism.

5. Conclusion

This research crafted a liposomal drug delivery system capable of concurrently encapsulating PMB and Fe²⁺. This innovative design cleverly circumvents the conventional PMB bactericidal route of direct LPS binding, instead harnessing the ferroptosis in bacteria, and subsequently induced an enhanced bactericidal action against PMB-resistant A. baumannii in vitro and in vivo. This work not only signifies a profound paradigm shift in antibiotic mechanisms of PMB but also rejuvenates the “aged” antibiotics by optimizing drug delivery rather than relying on novel chemical structures, emphasizing the potential applications of drug delivery technology in the battle with the drug-resistant bacteria.

CRediT authorship contribution statement

Xiangrong Wei: Writing – original draft, Methodology, Investigation, Formal analysis. Xinhui Cao: Writing – review & editing, Methodology, Formal analysis. Chengyi Xu: Methodology. Guangwei Shi: Methodology. Hong Wang: Formal analysis. Jinming Liu: Formal analysis. Huiyang Li: Supervision, Methodology. Bingmei Yao: Methodology. Yudong Zhang: Supervision. Liqun Jiang: Writing – review & editing, Writing – original draft, Supervision, Investigation, Conceptualization.

Declaration of competing interest

The authors declare that there are no conflicts of interest.

Appendix A. Supplementary data

The following is the Supplementary data to this article.