Multidrug-resistant (MDR) Escherichia coli (E. coli) infections pose a considerable threat to global public health, with devastating consequences to patient healthcare [1]. MDR Gram-negative infections pose the risk of making current antibiotic treatments ineffective due to the presence of two membranes (the inner and outer membranes) on the bacteria [2], [3] and the production of β-lactamases from bacteria [4]. Unfortunately, the discovery period for new antibiotics is 6-10 years long, whereas bacteria can evolve to tolerate 1000 times greater antibiotic concentrations than their wild-type ancestors within less than 2 weeks under laboratory conditions [5], [6]. More importantly, a recent study reported that MDR E. coli also showed resistance to the last-resort antibiotic colistin [7], [8], [9]. Ten million people per year will be killed due to a lack of effective antibiotics by 2050 [10]. Thus, it is urgent to develop an alternative strategy for effective and safe treatment against MDR E. coli.

Many strategies have been used in an attempt to address this issue, such as photoexcited quantum dots [11], probiotic-based nanoparticles (NPs) [12], and microwave-responsive Garcinia NPs [13]. These antibacterial strategies are mainly related to the generation of heat, reactive oxygen species (ROS), or antimicrobial peptides. However, they have a long way to go before clinical application. An ideal strategy would be to use US Food and Drug Administration (FDA)-approved materials to selectively kill pathogens while leaving normal cells undamaged. As one of the critical innate immune cells, macrophages (MΦs) play a vital role in preventing microbial invasion through ROS generated by biocatalytic process [14], [15], [16]. Importantly, MΦs only clear pathogens or apoptotic cells and do not attack normal cells, thanks to the receptors on the cell membrane. Considering their selectivity to normal cells and pathogens, MΦs have the potential for safely treating MDR E. coli infections. However, their physiological H₂O₂ concentrations (50-100 μmol∙L⁻¹) are too low for antibacterial activity. Fortunately, MΦs can be programmed by external stimuli due to their plastic phenotype [17], [18]. Pathogens and NPs (e.g., iron oxide NPs [19], CuS NPs [20], and Fe₃O₄@C/MnO₂-PGEA (PGEA is short for ethanolamine-functionalized poly(glycidyl methacrylate) [21]) can stimulate MΦs to produce H₂O₂. Among them, ferumoxytol and other iron oxide NPs have been approved by the US FDA [19]. Poly(lactic-co-glycolic acid) (PLGA) has also been approved by the US FDA [22]. Moreover, iron oxide NPs present dual enzyme-like activity both in vitro and in vivo, as they can react with H₂O₂ to form highly toxic ROS—namely, hydroxyl radicals (·OH)—in an acidic environment [23]. The bacterial infectious microenvironment has a low (i.e., acidic) pH [24]. Thus, MΦ-loaded Fe₃O₄@PLGA (MΦ-Fe₃O₄@PLGA) particles have behavior of intelligent catalysis, which could happen Fenton reaction to kill pathogen but without harming normal cells.

Here, we introduce an intelligent catalysis particle that kills MDR E. coli in a controlled manner without harming normal cells, by integrating an intelligent control center with a Fenton reaction catalyst. As shown in Fig. 1, this platform takes advantage of MΦ-Fe₃O₄@PLGA particles to produce H₂O₂ and lipid droplets (LDs) under the stimulation of MDR E. coli. These processes trigger the pathways of energy metabolism, infectious disease (bacterial), glycan biosynthesis and metabolism, and lipid metabolism. Moreover, the H₂O₂ produced further reacts with the Fe₃O₄ NPs to produce highly toxic ·OH. Furthermore, the LDs contain many antipathogenic proteins, which can target MDR E. coli. The antibacterial efficiency of the MΦ-Fe₃O₄@PLGA particles against MDR E. coli is shown to be 99.29% ± 0.31% in vitro. The MΦ-Fe₃O₄@PLGA particles are also activated in vivo and exhibit an excellent treatment effect on peritonitis in vivo.

Iron chloride hexahydrate (FeCl₃), dimethyl sulfoxide (DMSO), absolute ethanol (CH₃CH₂OH), ethylene glycol ((CH₂OH)₂), and sodium acetate (CH₃COONa) were obtained from Sinopharm Chemical Reagent Co., Ltd. (China). Sodium citrate (C₆H₅Na₃O₇), β-glycerol phosphate, dexamethasone, L-ascorbic acid, 3-[5-dimethylthiazol-2-yl]-2,5-diphenyl tetrazolium bromide (MTT), and PLGA (lactide:glycolide (50:50), molecular weight 30 000-60 000) were purchased from Sigma Chemical Co. (USA). Trypsin-EDTA, penicillin and streptomycin, fluorescein-5-isothiocyanate (FITC)-conjugated phalloidin (actin), and 4′,6-diamidino-2-phenylindole (DAPI) were obtained from Yeasen (China). Fetal bovine serum (FBS) was obtained from Gibco (US). An alkaline phosphatase (AKP) assay kit (microplate test kit) was purchased from Nanjing Jiancheng Bioengineering Institute (China). A BCA protein assay kit was purchased from Solarbio (China). 2′,7′-Dichlorofluorescein diacetate (DCFH-DA) was purchased from the Beyotime Institute of Biotechnology (China). PrimeScript RT Master Mix and 2× SYBR Premix Ex Taq II were obtained from TaKaRa (China).

Raw 264.7, L929, NIH-3T3, A549, MC3T3-E1, and bone marrow mesenchymal stem cells (BMSCs; two passages) were obtained from Nankai University (China) and cultured in growth mediums. The L929, NIH-3T3, A549, MC3T3-E1, and BMSCs were cultured in 89% (v/v) basic medium, 10% (v/v) FBS (OriCell), and 1% (v/v) penicillin-streptomycin (10 000 U·mL⁻¹, Gibco). For the L929 and NIH-3T3 cells, the basic medium was Dulbecco’s modified Eagle’s medium (DMEM; Biosharp, China). For the A549 cells and BMSCs, the basic medium was DMEM/Nutrient Mixture F-12 and α-minimum essential medium (α-MEM), respectively. MC3T3-E1 also used α-MEM as basic medium. The growth medium of the Raw 264.7 cells, MΦ, was 90% (v/v) DMEM and 10% (v/v) FBS.

MDR E. coli (China Center of Industrial Culture Collection (CICC), 10663) was grown in Luria-Bertani (LB) broth at 37 °C. MDR E. coli is resistant to the following antibiotics: gentamicin β-lactam antibiotics, ceftriaxone antibiotics, sulfisoxazole antibiotics, trimethoprim antibiotics, tetracycline antibiotics, and amoxicillin antibiotics.

First, 1.62 g of FeCl₃ and 7.2 g of anhydrous sodium acetate (CH₃COONa) were dissolved in 100 mL of (CH₂OH)₂ and underwent magnetic stirring for 3 h. Next, the solution was sealed in Teflon-lined stainless steel at 200 °C for 8 h to obtain solid particles. After this, the black solid particles were washed with CH₃CH₂OH three times (7000 r∙min⁻¹, 3 min) and dried in a vacuum for 24 h at 60 °C. These particles are referred to herein as Fe₃O₄ NPs.

The synthesis of the Fe₃O₄/PLGA NPs consisted of several steps, which were as follows. First, PLGA was dissolved in DMSO to make 1% (w/v) PLGA solution. Second, the PLGA solution was mixed with Fe₃O₄ solution (0.5 mg∙mL⁻¹ in deionized water) at a volume ratio of 1 to 100 under ultrasonic assistance and then cleaned with CH₃CH₂OH after centrifugation for 5 min at 7000 r∙min⁻¹. The final samples are referred to herein as Fe₃O₄@PLGA NPs.

To fabricate the MΦ-Fe₃O₄@PLGA particles, 2 mg of Fe₃O₄@PLGA NPs and 1 mL of Raw 264.7 (MΦs, 1 × 10⁶ cells per milliliter of DMEM) were co-cultured at 4 °C for 1 h to allow MΦs to uptake the Fe₃O₄@PLGA NPs to form MΦ-Fe₃O₄@PLGA particles. Finally, the MΦ-Fe₃O₄@PLGA particles were washed three times with DMEM.

FITC-Fe₃O₄@PLGA NPs were synthesized through a modified method. First, 1 mL of FITC (0.1 μmol∙L⁻¹) was dissolved in methanol, and 10 mg of Fe₃O₄@PLGA NPs was dispersed in 4 mL of methanol. Then, these solutions were thoroughly mixed via ultrasound. Next, the mixture was added to a pure 1-tetradecanol drop by drop. After stirring at 90 °C for 2 h, the methanol was evaporated. Next, the mixture was centrifuged at 10 000 r∙min⁻¹ for 3 min and washed with ultrapure water three times to obtain FITC-Fe₃O₄@PLGA NPs.

X-ray diffractometry (XRD; D8 Advanced, Bruker, Germany) using Cu Kα radiation was used to determine the crystal structure of the Fe₃O₄ and Fe₃O₄/PLGA NPs. Fourier-transform infrared spectroscopy (FTIR; Nicolet IS 10, Thermo Fisher Scientific Inc., USA) was performed to analyze the compositions of the different samples. Scanning electron microscopy (SEM; S4800, Hitachi High-Tech Corporation, Japan) was used to characterize the morphologies of the different samples. Transmission electron microscopy (TEM; FEI-Tecnai G2 Spirit TWIN and FEI-Talos F200X, FEI Company, USA) was used to obtain images of the samples.

Culture media from different groups (MΦs alone, co-cultured MΦs with MDR E. coli, MΦ-Fe₃O₄@PLGA particles, and co-cultured MΦ-Fe₃O₄@PLGA particles with MDR E. coli) was collected for the detection of ROS. Hydroxyl radicals were characterized by means of 3′-(p-hydroxyphenyl) fluorescein (HPF; 10 mmol∙L⁻¹) at 37 °C for 30 min. Finally, the optical density was obtained using an emission wavelength of 515 nm (excitation wavelength: 490 nm). H₂O₂ was measured using a catalase (CAT) assay kit (visible light, Nanjing Jiancheng Bioengineering Institute, China).

The cell-permeable fluorogenic probe DCFH-DA (10 μmol∙L⁻¹) was used to determine the content of intracellular ROS. Cellular esterases deacetylated the DCFH-DA into non-fluorescent DCFH after the DCFH-DA diffused into the cells. Then, the DCFH was oxidized to fluorescent 2′,7′-dichlorofluorescein in the presence of ROS. First, bacterial solution (10⁶ colony forming units per millilitre (CFU∙mL⁻¹)) was added to the MΦ-Fe₃O₄@PLGA group and cultured at 37 °C for 4 h. Next, the DCFH-DA was added to different samples (MΦ, MΦ-Fe₃O₄@PLGA particles, and MΦ-Fe₃O₄@PLGA particles with E. coli) and cultured for 30 min at 37 °C. Finally, an inverted fluorescence microscope (IFM; Olympus IX73, Olympus Corporation, Japan) was used to obtain the images.

A mitochondrial membrane potential assay kit (JC-1; SBJbio Life Sciences, China) was used to determine the mitochondrial membrane potential of the MΦs. ATP activity data was acquired using an ATP assay kit (Beyotime Biotechnology).

Bacterial cells were diluted to 2 × 10⁶ mL⁻¹ in each well with DMEM (pH = 6.0). After culturing with different samples for 4 h, the 20 μL liquid was spread on agar plates. Next, a glass spreader was used to spread the diluted liquid on the agar surface. Finally, the number of CFU was counted as CFU∙mL⁻¹.

To determine the antibacterial concentration of the Fe₃O₄@PLGA NPs, different concentrations of Fe₃O₄@PLGA NPs (1, 2, 4, and 8 mg·mL⁻¹) were cultured with 2 × 10⁶ CFU∙mL⁻¹ for 4 h, respectively. Finally, the number of bacteria was assessed via spread plate.

For transwell assays, dual-chamber transwell systems with 70 μm-sized microporous membranes were used to co-culture MC3T3-E1 cells and different groups (control, Fe₃O₄/PLGA NPs, MΦ, and MΦ-Fe₃O₄@PLGA). After co-culturing for 16 h, an MTT assay was performed on these cells; detailed steps were provided in our previous work [25].

To determine the toxicity of LDs to various normal cells (Raw 264.7, L929, NIH-3T3, A549, and BMSCs), lipopolysaccharides (LPS; 500 ng∙mL⁻¹) were used to treat these cells. After co-culturing for 16 h, an MTT assay was performed on these cells; detailed steps were provided in our previous work [25].

To assess the influence of the concentration of Fe₃O₄@PLGA NPs on cell viability, different concentrations of Fe₃O₄@PLGA NPs (1, 2, 4, 8, 16, and 32 mg∙mL⁻¹) were respectively cultured with 10⁴ cells per well for 24 h. Finally, the cell viability was analyzed via MTT assay.

For cell live/dead staining, 200 mL of solution (2 mg·mL⁻¹ Fe₃O₄ NPs or 2 mg·mL⁻¹ Fe₃O₄/PLGA NPs mixed with 10⁶ mL⁻¹, respectively) was co-cultured at 4 °C for 1 h. Next, the medium was removed, and 200 mL of phosphate-buffered saline (PBS) containing fluorescent dyes (1 μmol∙L⁻¹ calcein-AM (Ca-AM) and 10 μg·mL⁻¹ propidium iodide (PI)) was added. Next, the MΦs were washed three times with PBS after incubating for 30 min. Finally, images were obtained with an IFM.

To determine the distribution of the Fe₃O₄/PLGA NPs in MΦ-Fe₃O₄@PLGA, Raw 264.7 (10⁶ cells) was collected by centrifuging at 2000 r∙min⁻¹ for 3 min. The obtained Raw 264.7 was then co-cultured with 2 mg of FITC-Fe₃O₄@PLGA NPs at 4 °C for 1 h. Next, the mixture was centrifuged at 2000 r∙min⁻¹ for 3 min and then washed with DMEM three times. Then, the MΦ-FITC-Fe₃O₄@PLGA was fixed in 4% formaldehyde solution. After washing with PBS solution three times, the actin was stained with tetramethylrhodamine (TRITC)-conjugated phalloidin. Finally, images were obtained via laser scanning confocal microscopy (Nikon A1R+, Nikon, Japan).

For lysosome staining, Raw 264.7 (10⁶ cells) was collected by centrifuging at 2000 r∙min⁻¹ for 3 min. The Raw 264.7 was then co-cultured with 2 mg of FITC-Fe₃O₄@PLGA NPs at 4 °C for 1 h. Next, the lysosome was cultured with LysoTracker Red for 20 min. Then, the medium was removed and fresh growth medium was added. Finally, images were obtained via laser scanning confocal microscopy.

TRIzol reagent (Invitrogen, Thermo Fisher Scientific Inc.) was used to extract the total RNA. Nanodrop 2000 and agarose gel electrophoresis were performed to determine the concentration and purity of the RNA and to assess RNA integrity. The data were analyzed on the Majorbio Cloud Platform.

RSEM (version 1.3.1, GitHub Inc., USA) was used to assess the relationship between these samples. The false discovery rate method was corrected for P value. DESeq2, DEGseq, and edgeR were used to obtain a gene differential expression analysis, which was further applied to gene ontology (GO) and the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway. The GO and KEGG pathway analyses were processed using Fisher’s exact test and the χ² test.

BALb/c mice were used for in vivo antibacterial experiments with a peritonitis model. Animal testing was performed following the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The Animal Ethical and Welfare Committee (AEWC) of the Institute of Radiation Medicine, Chinese Academy of Medical Sciences, approved the ethical part of the experiment. A total of 48 male BALb/c mice weighing about 20 g were divided into four groups. The DMEM, MΦ, Fe₃O₄@PLGA, and MΦ-Fe₃O₄@PLGA groups were respectively injected with 100 μL of DMEM solution containing 10⁸ CFU of E. coli, 100 μL of DMEM solution containing 2 × 10⁶ MΦ cells and 10⁸ CFU E. coli, 100 μL of DMEM solution containing 2 mg of Fe₃O₄@PLGA NPs and 10⁸ CFU E. coli, or 100 μL of DMEM solution containing 100 μL of MΦ-Fe₃O₄@PLGA particles (2 × 10⁶ cells-2 mg Fe₃O₄@PLGA NPs) containing 10⁸ CFU E. coli. Finally, a histopathological evaluation and blood routine assay were performed at the appropriate time.

The peritoneal lavage was collected with cold PBS solution. Seventy-micrometer nylon strainers were used to prepare single-cell suspensions. Then, Fc receptor binding inhibitor diluted 1:10 in PBS containing Zombie-Aqua fixable viability stain was used to exclude dead cells by co-culturing with the cells for 15 min on ice. The cells were then incubated according to the manufacturers’ instructions using dilutions of fluorescently labeled primary monoclonal antibodies. The following antibodies were used: PerCP/Cyanine5.5 anti-mouse CD11c and PE anti-mouse CD206 antibody. Data were obtained by means of an Agilent flow cytometer (NovoCyte 2000, ACEA Biosciences, Inc., USA).

Data are presented as mean ± standard deviation (SD). Paired or unpaired t-tests, one-way analysis of variance (ANOVA), and two-way ANOVA were used in the statistical analyses. P < 0.05 was considered statistically significant.

The design of the MΦ-Fe₃O₄@PLGA particles relied on the phagocytosis of the MΦs of the Fe₃O₄@PLGA NPs. The fabrication process is illustrated in Fig. 2(a). First, 0.01% (w/v) PLGA solution was used to modify the Fe₃O₄ NPs to form Fe₃O₄@PLGA NPs. Next, the MΦs were collected using trypsin and co-cultured with the Fe₃O₄@PLGA NPs for 1 h at 4 °C in DMEM. Subsequently, DMEM was used to wash the above particles three times to eliminate unbound cells. Thus, the MΦ-Fe₃O₄@PLGA particles were formed. SEM was used to characterize the morphology of the Fe₃O₄ NPs and Fe₃O₄@PLGA NPs (Figs. S1 and S2 in Appendix A). The Fe₃O₄ NPs, which had a sphere-like morphology, were nearly (318.00 ± 100.66) nm in diameter (Fig. S1). As for the Fe₃O₄@PLGA NPs, the PLGA did not change the morphology of the Fe₃O₄ NPs (Fig. S2, left). Elemental mapping images showed that the carbon element was evenly distributed on the Fe₃O₄ NPs, further demonstrating that the PLGA was coated onto the Fe₃O₄ NPs (Fig. S2, right). The crystal structure of Fe₃O₄ and the Fe₃O₄@PLGA NPs was also determined from the XRD pattern (Fig. 2(b)). The peaks at (220), (311), (400), (422), (511), (440), and (533) are characteristic diffraction peaks of the Fe₃O₄ NPs [16], [26]. FTIR was further used to characterize the PLGA, and strong absorption peaks at 2966, 2937, 1748, and 1078 cm⁻¹ were ascribed to the C-H stretch of CH₂, the C-H stretch of -C-H-, the stretching vibration of C=O, and the C-O stretching, respectively (Fig. 2(c)). These peaks are the characteristic peaks of the PLGA molecule [27]. The results indicated that the Fe₃O₄@PLGA NPs were successfully formed. Next, MΦ-Fe₃O₄@PLGA particles were formed, following the method illustrated in Fig. 2(a). Fig. 2(d) displayed an image of the MΦ-Fe₃O₄@PLGA particles, where the MΦs and Fe₃O₄@PLGA NPs were blue and peach in color, respectively. Fig. 2(d) (left) showed that some Fe₃O₄@PLGA NPs were bound on the surface of the MΦ. Fig. S3 (in Appendix A) also demonstrated a similar phenomenon. Fig. 2(d) (right) further demonstrated that the morphology of the Fe₃O₄@PLGA NPs did not change on the MΦ surface. TEM images were additionally used to assess the distribution of the Fe₃O₄@PLGA in the MΦ-Fe₃O₄@PLGA particles; they showed that the Fe₃O₄@PLGA NPs (marked by blue arrows) were associated both extracellularly and intracellularly with the MΦs (Fig. 2(e), left). The high-resolution TEM (HRTEM) image (Fig. 2(e), right) showed that the adjacent lattice fringes were approximately 0.48 nm, corresponding to the [111] plane of the Fe₃O₄ NPs with a cubic inverse spinel structure [28].

The zeta potential of different samples was measured, and the zeta potentials of the MΦs, Fe₃O₄ NPs, Fe₃O₄@PLGA NPs, and MΦ-Fe₃O₄@PLGA particles were found to be −24.93, −3.37, −0.31, and −10.43 mV, respectively (Fig. 2(f)). The cell membrane is negatively charged due to the phospholipid bilayer structure. Negatively charged superparamagnetic particles have been shown to exhibit a high but nonspecific affinity for the cell membrane [29]. Some Fe₃O₄@PLGA NPs were distributed on the surface of the cell membrane, and the zeta potential was increased from −24.93 to −10.43 mV. These results showed that the Fe₃O₄@PLGA NPs were successfully connected with the MΦs.

Ca-AM (green) and PI (red) were used to stain viable and dead cells. As shown in Fig. 2(g), the color of the cells was red after co-culturing the Fe₃O₄ NPs with MΦs at 4 °C for 1 h. In contrast, the cells were green after co-culturing with the Fe₃O₄@PLGA NPs. These results indicated that PLGA had a remarkable protective effect on the MΦs during the synthesis process of the MΦ-Fe₃O₄@PLGA particles. Previously, it has been demonstrated that Fe₃O₄ NPs could increase the inflammatory response of MΦs [30]. Moreover, naked superparamagnetic iron oxide NPs were found to present obvious cytotoxicity [31], [32]. PLGA has a low propensity to cause immune responses and is not toxic to cells [30], [31]. The Fe₃O₄ NPs were coated with PLGA, resulting in low toxicity with little or no effect on cell function and viability [33]. It should be noted that a cold experimental setting can up-regulate the expression of pro-inflammatory genes [34] and limit the phagocytic behavior of MΦ.

ATP is generally considered to be the “energy currency” of the cell [35]. The production of ATP by the MΦs in the MΦ-Fe₃O₄@PLGA particles was obviously greater than that produced by the MΦs alone (Fig. 2(h)). When MΦs are activated into M1 MΦs, their metabolism switches from oxidative phosphorylation to aerobic glycolysis—a change that has been linked to the generation of ROS [36], [37]. The enhanced aerobic glycolysis can rapidly provide ATP from glucose, enhancing the ATP activity. The results indicated that the viability and function of the MΦ-Fe₃O₄@PLGA particles was not damaged by the synthesis process, and that Fe₃O₄@PLGA had the potential to induce MΦs into the M1 phenotype.

The cytotoxicity of the MΦ-Fe₃O₄@PLGA particles on normal cells was further assessed as shown in Fig. 2(i), with MC3T3-E1 being chosen as a model cell. Compared with the cell viability of MC3T3-E1 cultured with DMEM, MC3T3-E1 cultured with the MΦ-Fe₃O₄@PLGA and Fe₃O₄@PLGA groups showed slightly higher cell viability. These results suggested that the Fe₃O₄@PLGA NPs and MΦ-Fe₃O₄@PLGA particles showed excellent biocompatibility with MC3T3-E1. To further investigate the influence of the MΦ-Fe₃O₄@PLGA particles on the viability of MDR E. coli, the antibacterial activity of different samples (DMEM, MΦ, Fe₃O₄@PLGA, and MΦ-Fe₃O₄@PLGA) was assessed (Figs. 2(j) and (k)). Fe₃O₄@PLGA NPs with concentration of 4 mg∙mL⁻¹ were found to be fatal to bacteria without detriment to the MΦ by means of MTT and antibacterial assays (Figs. S4 and S5 in Appendix A). Compared with the CFU of MDR E. coli treated with DMEM, the CFU of MDR E. coli in MΦ, Fe₃O₄@PLGA, and MΦ-Fe₃O₄@PLGA were 1.37-log, 0.75-log, and 2.33-log decreases, respectively. The antibacterial efficiency against MDR E. coli of MΦ, Fe₃O₄@PLGA, and MΦ-Fe₃O₄@PLGA was found to be 95.74% ± 0.44%, 82.22% ± 0.95%, and 99.29% ± 0.31%, respectively (Figs. 2(j) and (k)). These results demonstrated that the MΦ-Fe₃O₄@PLGA particles could selectively kill MDR E. coli without harming normal cells.

As mentioned in the earlier section on the intelligent-killing behavior of the MΦ-Fe₃O₄@PLGA particles, the MΦ-Fe₃O₄@PLGA particles were chosen for an in vitro study. Next, we further analyzed the influence of MDR E. coli on the MΦ’s fate in MΦ-Fe₃O₄@PLGA or MΦ. The polarization phenotypes of the MΦs in the MΦ-Fe₃O₄@PLGA particles were characterized by means of enzyme-linked immunosorbent assay (ELISA) (Fig. 3(a)). Tumor necrosis factor-α (TNF-α) and interleukin (IL)-10 were chosen as a classical M1 marker [38] and M2 marker [39], respectively. As shown in Fig. 3(a), compared with the MΦs alone, the MΦs of the MΦ-Fe₃O₄@PLGA particles showed similar expression of the cytokines of IL-10 and TNF-α, suggesting that the Fe₃O₄@PLGA NPs did not cause MΦ polarization. In the MΦs treated with E. coli, the production of TNF-α was increased, but there was no significant production of IL-10, indicating that E. coli was able to promote the MΦs into the M1 phenotype. Unexpectedly, the concentration of TNF-α was even higher after co-culturing E. coli with the MΦ-Fe₃O₄@PLGA particles. This result indicated that E. coli promoted M1 MΦ polarization. Fluorescence-activated cell sorting (FACS) was further performed to analyze the influence of MDR E. coli on the polarization phenotypes of MΦ-Fe₃O₄@PLGA (Figs. 3(b) and (c)). Compared with the untreated MΦs, 2.63% and 3.55% of the MΦs of the MΦ-Fe₃O₄@PLGA and of the MΦ-Fe₃O₄@PLGA with E. coli responded to the M1 marker (CD11c), respectively, and 1.31% and 0.60% responded to the M2 marker (CD206) [38]. The ratio between M1 and M2 in the MΦ-Fe₃O₄@PLGA and MΦ-Fe₃O₄@PLGA with E. coli groups was 2.01 and 5.92, respectively, indicating that E. coli further induced M1 MΦ polarization in the MΦ-Fe₃O₄@PLGA particles. These results demonstrated that the MΦ-Fe₃O₄@PLGA was able to polarize into an M1 MΦ after treatment with MDR E. coli.

It is generally known that M1 MΦs release H₂O₂ [40]. H₂O₂ is an essential component of the Fenton reaction, which generates highly toxic hydroxyl radicals in the presence of iron [19]. To further assess the possibility of the Fenton reaction occurring, H₂O₂ was externally supplied to assess whether the Fe₃O₄@PLGA NPs could react with H₂O₂ to produce hydroxyl radicals. An HPF detection kit was used to detect hydroxyl radicals. As shown in Fig. 3(d), the produced hydroxyl radicals were positively related to the content of the supplied H₂O₂ (0, 2, 8, 32, and 128 mmol∙L⁻¹). Thus, the Fe₃O₄@PLGA NPs reacted with the H₂O₂ supplied to produce hydroxyl radicals. Moreover, co-cultures of MDR E. coli and MΦ-Fe₃O₄@PLGA produced 138.39 mmol∙L⁻¹ H₂O₂ compared with MΦs alone (51.74 mmol∙L⁻¹ H₂O₂; P < 0.0001) (Fig. 3(e)), which was a 4.80-fold increase in hydroxyl radicals compared with the MΦs (P < 0.0001) (Fig. 3(f)).

Next, we investigated whether co-cultures of MDR E. coli could enhance the intracellular accumulation of soluble iron species. The intracellular accumulation of soluble iron species of different samples was detected by means of inductively coupled plasma (ICP), as shown in Fig. 3(g). Compared with the MΦ group, the level of intracellular soluble iron was dramatically higher in the cells from Fe₃O₄@PLGA particles. This was mainly due to the presence of the Fe₃O₄@PLGA NPs. Moreover, adding MDR E. coli further enhanced the concentration of iron. This phenomenon indicated that the MΦ-Fe₃O₄@PLGA particles produced soluble iron species in response to MDR E. coli. TEM images were used to display the presence of LDs in different samples (MΦ, MΦ-Fe₃O₄@PLGA, and MΦ-Fe₃O₄@PLGA with E. coli) (Fig. S6 in Appendix A; Figs. 3(h) and (i)). Compared with the MΦs, there was no presence of LDs in the MΦ-Fe₃O₄@PLGA particles. After co-culturing with MDR E. coli, some LDs (red arrows) and MDR E. coli (green arrows) were present in the cells. Some LDs were located around the MDR E. coli (green arrows), demonstrating that the LDs could target MDR E. coli. LDs are major lipid storage organelles of eukaryotic cells; they contain antibacterial proteins that participate in antibacterial processes [41]. It is generally known that LDs have protein-mediated antimicrobial capacity, and infection can increase the generation of LDs. This result indicated that LDs were formed in the MΦ-Fe₃O₄@PLGA particles after treatment with MDR E. coli. It is generally known that LPS can induce LD formation in cells [41]. LPS were used to treat the Raw 264.7, L929, A549, NIH-3T3, and BMSCs for 16 h. As shown in Fig. S7 (in Appendix A), LPS had no toxic effect on these cells. The results suggested that LDs would not kill normal cells.

The content of intracellular ROS was detected by means of DCFH-DA [42], [43]. The green fluorescence intensity of the MΦ-Fe₃O₄@PLGA particles was higher than that of the MΦs, suggesting that the content of intracellular ROS was enhanced in response to the Fe₃O₄@PLGA NPs. After MDR E. coli was added, the green fluorescence intensity was further enhanced, suggesting that MDR E. coli promoted the content of intracellular ROS (Fig. 3(j)). The corresponding quantitative analysis showed a similar tendency (Fig. 3(k)). These results further confirmed that the MΦ-Fe₃O₄@PLGA particles produced more ROS in response to MDR E. coli. Live cell imaging was further used to characterize the location of the ROS. The cell membrane, lysosome, intracellular ROS, and nucleus were stained with DiR, LysoTracker Red, a DCFH-DA probe, and Hoechst 33342, respectively. As shown in Fig. S8 in Appendix A, green fluorescence appeared to overlap with the red fluorescence and also existed in other areas, suggesting that ROS were produced in parts of the cell other than lysosomes. Fig. S9 (in Appendix A) further indicates that ROS were not only produced in the cells but also in the environment. FITC was used to mark the Fe₃O₄@PLGA NPs, and TRITC-labeled phalloidin was used to stain the actin of the MΦs. As shown in Fig. 3(l), the Fe₃O₄@PLGA NPs were distributed both on the membrane and in the intracellular system of the MΦs. Lysosome staining was further performed to assess whether Fe₃O₄@PLGA could enter lysosomes. As shown in Fig. 3(m), some Fe₃O₄@PLGA NPs were distributed in lysosomes. This result indicated that some of the Fe₃O₄@PLGA NPs were engulfed by lysosomes. PLGA is negatively charged, and the negatively charged NPs prefer to colocalize with endosomes and lysosomes [44]. These results demonstrated that the MΦ-Fe₃O₄@PLGA particles had intelligent catalytic ability and could activate a selective Fenton reaction in response to MDR E. coli in lysosomes and the infectious microenvironment.

To further analyze the underlying mechanism of the intelligent catalysis behavior of the MΦ-Fe₃O₄@PLGA particles, high-throughput sequencing was used to analyze gene expression profiles [25]. RNA-sequence analysis was performed to study the expression difference of the MΦs under different conditions (MΦs alone, MΦ with E. coli, and MΦ-Fe₃O₄@PLGA with E. coli). A total of 2030 genes were detected: 631 genes in MΦ-Fe₃O₄@PLGA with E. coli versus MΦs, 441 genes in MΦ-Fe₃O₄@PLGA with E. coli versus MΦs with E. coli, and 52 genes in MΦs with E. coli versus MΦs (Fig. 4(a)). Principal component analysis (PCA) revealed distances between different samples (MΦs, MΦs with E. coli, and MΦ-Fe₃O₄@PLGA with E. coli), suggesting that the different treatments would cause differences in gene expression (Fig. 4(b)). As shown in Figs. 4(c)-(e), the volcano plots showed 273 up-regulated genes and 43 down-regulated genes for MΦs with E. coli versus MΦs; 644 up-regulated genes and 718 down-regulated genes for MΦ-Fe₃O₄@PLGA with E. coli versus MΦs; and 408 up-regulated genes and 900 down-regulated genes for MΦ-Fe₃O₄@PLGA with E. coli versus MΦs with E. coli. This finding suggested that the treatment caused a major difference in gene expression.

A GO database analysis was performed to analyze the expression of different genes. The enriched terms of MΦ-Fe₃O₄@PLGA with E. coli versus MΦs with E. coli, and those of MΦs with E. coli versus MΦs, are shown in Fig. 4(f). The genes were rich in catalytic activity, response to stimulus, biological adhesion, and immune system process on MΦs with E. coli versus MΦs. The genes were rich in catalytic activity, antioxidant activity, response to stimulus, biological adhesion, and immune system process on MΦ-Fe₃O₄@PLGA with E. coli versus MΦs with E. coli. According to the KEGG pathway analysis, it was found that infectious disease (bacterial), immune system, lipid metabolism, and glycan biosynthesis and metabolism were up-regulated in MΦs with E. coli compared with MΦs (Fig. S10 in Appendix A). Infectious disease (bacterial), lipid metabolism, glycan biosynthesis and metabolism, and energy metabolism were up-regulated in MΦ-Fe₃O₄@PLGA with E. coli compared with MΦs with E. coli (Fig. S11 in Appendix A). These activated signaling pathways and functions resulted in M1 MΦ development and LD formation.

From the heatmap of MΦ-Fe₃O₄@PLGA with E. coli, MΦs with E. coli, and MΦs, the expression of the genes chemokine (C-C motif) ligand 9 (CCL9) [45], interleukin 6 (IL-6) [46], complement component 3 (C3) [47], schlafen 4 (SLFN4) [48], abhydrolase domain containing 3 (ABHD3) [49], abhydrolase domain containing 1 (ABHD1) [50], cluster of differentiation antigen (CD), and transforming growth factor β regulator 4 (TBRG4) [51] were changed (Fig. 4(g)). The genes of IL-6, CCL9, SLFN4, and C3 were related to the polarization of MΦs. Moreover, the genes of ABHD1, ABHD3, and CD were related to the formation of LDs in MΦs.

Mitochondria play a vital role in immunity response, and the interaction between LDs and mitochondria is decreased in infected cells [41]. The mitochondria membrane potential is a signal of mitochondria activity [33]. As shown in Fig. S12 (in Appendix A), the ratio of green/red fluorescence intensity in MΦs with E. coli was higher compared with the MΦ group, and that of the MΦ-Fe₃O₄@PLGA with E. coli group was lower compared with the MΦs with E. coli group, suggesting that the interaction between LDs and mitochondria was decreased, and that more LDs were involved in antibacterial activity. Based on the above results, the complete process of clearing MDR E. coli via MΦ-Fe₃O₄@PLGA is illustrated in Fig. 4(h). Iron oxide NPs are phagocytized by MΦs and are then degraded into iron ions within lysosomes [52]. When MΦ-Fe₃O₄@PLGA particles are co-cultured with MDR E. coli, the expression of the genes TNF-α, IL-6, CCL9, SLFN4, and C3 is up-regulated, while the expression of the gene TBRG4 is down-regulated. These results indicate that the MΦs are polarized into the M1 phenotype. The M1 MΦs release hydrogen peroxide, which generates highly toxic hydroxyl radicals (·OH) in the presence of iron ions via the Fenton reaction [53]. Furthermore, the genes of ABHD3, ABHD1, and CD were up-regulated, indicating that LDs were formed in the MΦs in response to MDR E. coli.

The hydroxyl radicals and LDs finally caused the death of the MDR E. coli. The viability and function of the MΦ-Fe₃O₄@PLGA particles were further evaluated by means of Ca-AM/PI staining and an antibacterial assay (Figs. 5(a) and (b)). The number of remaining cells in the MΦ-Fe₃O₄@PLGA group decreased after several passages (Fig. 5(a)). The antibacterial efficiency of MΦ-Fe₃O₄@PLGA at passages 1 and 2 was 62.04% ± 3.84% and 17.00% ± 5.44%, respectively (Fig. 5(b)). These results suggested that the function of MΦ-Fe₃O₄@PLGA was partially retained after several passages.

The aforementioned in vitro investigation demonstrated that the MΦ-Fe₃O₄@PLGA particles had excellent selectivity between pathogens and normal cells. The MΦ-Fe₃O₄@PLGA particles were therefore selected for an in vivo study. Fluorescence imaging was performed to evaluate whether the MΦ-Fe₃O₄@PLGA particles could be retained and accumulate at the infected site for enough time in vivo (Figs. 5(c)-(e)). MΦ-Fe₃O₄@PLGA-Cyanine-7 (cy7) was used for this process, and images at different time points were obtained to display the process. Compared with the pre-injection group (mice without any treatment), a significantly high fluorescence signal was shown at 0, 2, and 6 h for mice treated with MΦ-Fe₃O₄@PLGA-cy7. Moreover, the fluorescence disappeared after 24 h. This result indicated that MΦ-Fe₃O₄@PLGA was able to remain at the infected site for at least 6 h.

Next, fluorescence images of major organs (i.e., the liver, lung, heart, spleen, and kidney) were further used to analyze the metabolic behavior of MΦ-Fe₃O₄@PLGA. The fluorescence images of major organs showed that some MΦ-Fe₃O₄@PLGA-cy7 particles were distributed in the liver and kidney at 6 h. The imaging data demonstrated that MΦ-Fe₃O₄@PLGA could be metabolized by the liver and kidney. Furthermore, fluorescence was absent after 18 h, indicating that the MΦ-Fe₃O₄@PLGA particles were completely metabolized. The cytotoxicity of MΦ-Fe₃O₄@PLGA was then further analyzed in vivo. The liver and kidney function of each group were assessed to test biosafety in vivo. No difference was found between the groups, suggesting that MΦ-Fe₃O₄@PLGA was not toxic to the liver and kidney (Figs. 5(f) and (g)). Hematoxylin and eosin (H&E) staining of the major organs was performed to further characterize the toxicity of MΦ-Fe₃O₄@PLGA to major organs. The control and MΦ-Fe₃O₄@PLGA groups showed no obvious damage to the major organs (Fig. 5(h)). These results indicated that MΦ-Fe₃O₄@PLGA has excellent biosafety in vivo.

To further assess whether the MΦ-Fe₃O₄@PLGA particles could treat peritonitis in vivo, a peritonitis model was constructed by injecting MDR E. coli into mice (Fig. 6(a)). The control group was treated with DMEM, while the other groups were respectively treated with MΦs, Fe₃O₄@PLGA, and MΦ-Fe₃O₄@PLGA particles. After 1 h of infection, the blood of different groups was collected to perform the blood chemistry and routine analysis. White blood cells (WBCs), lymphocytes (Lymph#), monocytes (Mon#), and granulocytes (Gran#) were chosen as the evaluation index for infection. These are immune cells that participate in fighting illness and disease and thus can be used to evaluate the immune response [54], [55]. As shown in Fig. 6(b), the values of WBC, Lymph#, Mon#, and Gran# in the MΦ-Fe₃O₄@PLGA group were lower than in the control, MΦs, and Fe₃O₄@PLGA groups. On the other hand, compared with the control group, the values in the MΦs and Fe₃O₄@PLGA groups were greater. These results indicated that the MΦ particles and Fe₃O₄@PLGA particles had a certain antibacterial effect in vivo, and that the MΦ-Fe₃O₄@PLGA particle had a strong antibacterial effect in vivo. The bacterial number in major organs from different groups (i.e., the control, MΦ, Fe₃O₄@PLGA, and MΦ-Fe₃O₄@PLGA groups) was characterized in Fig. 6(c). Compared with the control group, the CFU value of the MΦ, Fe₃O₄@PLGA, and MΦ-Fe₃O₄@PLGA groups were lower. Moreover, the CFU value of the MΦ-Fe₃O₄@PLGA group was even lower compared with the MΦ and Fe₃O₄@PLGA groups. These phenomena indicated that the MΦ-Fe₃O₄@PLGA particles had strong antibacterial ability and a certain protective effect on major organs in vivo. The bacterial value of the blood from different groups showed similar results (Fig. S13 in Appendix A). The phenotype of the MΦs in the abdominal cavity and blood was assessed via ELISA (Fig. 6(d) and Fig. S14 in Appendix A). Compared with the control group, the ratio between IL-10 (the M2 marker) and TNF-α (the M1 marker) in the MΦ-Fe₃O₄@PLGA group was increased. This result suggested that the MΦ-Fe₃O₄@PLGA group was anti-inflammatory. FACS showed that the ratio between the M1 phenotype (CD11c⁺CD206⁻) and the M2 phenotype (CD11c⁻CD206⁺) in the control, MΦ, Fe₃O₄@PLGA, and MΦ-Fe₃O₄@PLGA groups was 8.28, 26.16, 50.13, and 55.06, respectively (Fig. 6(e)). This result was mainly due to the decreased bacterial burden in the MΦ-Fe₃O₄@PLGA group in vivo.

The peritoneum tissues were harvested and sectioned for immunohistochemistry staining (Fig. S15 in Appendix A). The immunohistochemistry staining of the control, MΦ, Fe₃O₄@PLGA, and MΦ-Fe₃O₄@PLGA groups showed that the expression of TNF-α in MΦ-Fe₃O₄@PLGA was down-regulated, and the expression of IL-10 was increased (Fig. S15). Next, the H&E staining of major organs for the control, MΦ, Fe₃O₄@PLGA, and MΦ-Fe₃O₄@PLGA groups were performed at Day 2 (Fig. 6(f)). In the figure, inflammatory cells are marked by green arrows in the control, and Fe₃O₄@PLGA groups. In contrast, the MΦ-Fe₃O₄@PLGA group had almost no inflammatory cells. These results indicated that the MΦ-Fe₃O₄@PLGA particles had a strong treatment effect on MDR E. coli-induced peritonitis.

In summary, this paper reported on the development of MΦ-Fe₃O₄@PLGA particles as biomimetic intelligent catalysts. The MΦ-Fe₃O₄@PLGA particles demonstrated excellent antibacterial ability through biocatalysis and Fenton catalysis. The particles were polarized into the M1 phenotype under the stimulation of MDR E. coli. The M1-type MΦs produced H₂O₂ and LDs through biocatalysis. Selective Fenton catalysis occurred due to the presence of H₂O₂ in the lysosome and the infectious microenvironment. LDs targeted E. coli and participated in the antibacterial process. More importantly, the cells’ viability, integrity, and function were retained after several passages. PLGA and Fe₃O₄ NPs have been approved for use in humans by the US FDA. Overall, MΦ-Fe₃O₄@PLGA particles may become an “off-label” drug for clinical applications. Biomimetic intelligent catalysis takes full advantage of the properties of MΦs and Fe₃O₄@PLGA NPs to treat infections and can be extended to other cells and NPs to treat disease. Nevertheless, the culturing condition of living cells limits its clinical application.

This work is jointly supported by the China National Funds for Distinguished Young Scientists (51925104), the National Natural Science Foundation of China (NSFC) (51871162 and 52173251), NSFC-Guangdong Province Joint Program (U21A2084), and the Central Guidance on Local Science and Technology Development Fund of Hebei Province (226Z1303G).

Jieni Fu, Xiangmei Liu, Zhaoyang Li, Yufeng Zheng, Yu Zhang, Hui Jiang, Yanqin Liang, Shengli Zhu, Zhenduo Cui, and Shuilin Wu declare that they have no conflict of interest or financial conflicts to disclose.

Supplementary data to this article can be found online at https://doi.org/10.1016/j.eng.2023.05.022.