Engineered vesicles enhance oral antibiotic absorption in proximal small intestine and mitigate gut dysbiosis

Abstract

Oral antibiotics are a mainstay for treating bacterial infections, but unabsorbed portions can reach the caecum and colon, leading to gut dysbiosis. Herein, we engineer a biohybrid delivery system through the integration of milk extracellular vesicles with liposomes. The hybrid vesicles employ targeting mechanisms via neonatal Fc receptor and peptide transporter 1, facilitating antibiotic transport across the proximal small intestine. These vesicles exhibit superior drug encapsulation efficiency, stable release behavior, efficient mucus traversal, higher endocytosis, increased basolateral exocytosis, and improved oral absorption, achieving a 3.24-fold increase in oral bioavailability compared to free antibiotics. In lung bacterial infections and bacteremia models, hybrid vesicle-encapsulated cefdinir outperforms free antibiotics in eliminating infections. Notably, this approach also mitigates adverse effects on the intestinal microbiota, safeguarding the animals from dysbiosis-associated metabolic syndromes and opportunistic pathogen infections. This innovative hybrid vesicle system holds promise for the oral delivery of other drugs that suffer from limited absorption or cause gut dysbiosis.

In this work, authors present a biohybrid delivery system aimed at overcoming low oral bioavailability and antibiotic-induced dysbiosis by enabling efficient antibiotic transport in the proximal small intestine.

Introduction

Antibiotics are among the most widely prescribed medications globally, with their usage escalating at an alarming rate¹. However, the majority of antibiotics are hampered by their low oral bioavailability, necessitating the use of excessive doses to achieve the desired therapeutic outcomes. The frequent administration of high doses of antibiotics not only exerts a significant burden on the body but also fosters the emergence of antibiotic-resistant strains². Moreover, the unabsorbed fraction of orally administered antibiotics reaches the caecum and colon, potentially causing intestinal microbiota dysbiosis³. The mutualistic microbes residing in the intestine are vital for various physiological functions, including the regulation of immune and metabolic homoeostasis^4–6. When this delicate balance is disrupted by antibiotics, it may trigger a cascade of adverse effects, such as diarrhea, allergic reactions, inflammation, and neurological disorders^7–11. Thus, it is essential to improve the oral bioavailability of antibiotics and reduce their unintended negative impacts on gut microbiota to optimize their use.

The population and composition of gut microbiota differ greatly in various parts of the intestine. Bacterial counts typically stretch from approximately 10⁵/mL in the upper small intestine to approximately 10¹²/mL in the colon¹². This bacterial density gradient hints that if antibiotics are effectively absorbed in the proximal small intestine, it would restrict contact with the large intestine’s flora, thereby cutting down their presence in feces¹³. In recent years, extracellular vesicles have emerged as promising drug delivery carriers for targeted delivery and overcoming physiological barriers^14–16. Milk extracellular vesicles (mEV) stand out as particularly promising oral delivery candidates due to their ample sources, efficient cargo protection capabilities, minimal immunogenicity, and stable gastrointestinal physicochemical properties¹⁷. Notably, research has shown that mEV are transported through neonatal Fc receptor (FcRn) and peptide transporter 1 (PEPT1) on the intestinal epithelial cells¹⁸. Since FcRn and PEPT1 are highly expressed in the proximal small intestine, this transport mechanism may promote absorption by the proximal small intestine while minimizing side effects on the microbiota in the colon^19,20. Nevertheless, natural mEV still face several limitations as oral carriers. First, mEV have limited ability to traverse the mucus layer²¹. Second, mEV encounter challenges in encapsulating cargo efficiently²². Third, the phenomenon of rapid release of small molecule drugs encapsulated in mEV is quite pronounced²³. To overcome these limitations and construct a more efficient carrier for oral delivery, it is essential to modify natural mEV to enhance their mucus penetration ability, drug loading capacity, and release behavior.

In this study, we design a milk extracellular vesicle-liposome (mEV-Lip) hybrid vesicle with the aim of enhancing absorption in the proximal small intestine and thereby circumventing adverse effects on the colonic microbiota (Fig. 1). The hydrophilic modification of polyethylene glycol (PEG) enables the mEV-Lip to penetrate the intestinal mucus layer, while the proteins on the surface of mEV facilitate targeting to epithelial cells in the proximal small intestine, thus enhancing the targeted absorption of antibiotics. We select cefdinir (Cef), a β-lactam antibiotic, as the representative antibiotic for our study. The results demonstrate that oral administration of Cef-loaded mEV-Lip (mEV-Lip@Cef) exhibits higher efficacy in Klebsiella pneumoniae infection in both lung and bacteremia models. More importantly, treatment with mEV-Lip@Cef markedly diminishes the negative impact on intestinal microbiota, mitigating the risk of dysbiosis-associated opportunistic pathogen infection and metabolic syndromes.

Fig. 1. Schematic illustration of the milk extracellular vesicle-liposome hybrid vesicle (mEV-Lip) for oral delivery of antibiotics.

The mEV-Lip system is designed to enhance the absorption of antibiotics in the proximal small intestine, thereby reducing the contact between antibiotics and the microbiota residing in colon, thus mitigating antibiotic-associated dysbiosis.

Results

Hybrid vesicles facilitate enhanced absorption at proximal small intestine

The uptake of mEV at the intestinal epithelium is facilitated by FcRn and PEPT1¹⁸. To explore the potential of targeted antibiotic delivery, we first investigated the expression of these transporters in the intestine. As depicted in Fig. 2a and Supplementary Fig. 1, the expression levels of FcRn and PEPT1 were notably elevated in small intestine relative to the colon. This disparity in transporter distribution suggests a promising avenue for transporter-mediated targeting of mEV in the proximal small intestine.

Fig. 2. Hybrid vesicles enhance absorption in the proximal small intestine.

a Immunofluorescence analysis of FcRn and PEPT1 expression in mouse small intestines and colon. Scale bars, 100 μm. Experiments were repeated twice. b Dynamic light scattering-measured mEV particle size. c Nanoparticle-tracking analysis measurement of mEV. d FRET phenomenon of mEV-Lip encapsulating C6 and RhB. e WB results of mEV characteristic proteins. f Representative trajectories of different vesicles in fresh porcine mucus. Scale bars, 50 pixels. g The mean square displacement (MSD) of vesicles over time. h Quantitative exocytosis of various vesicles. ****P = 0.0000014 for the comparison between the mEV group and mEV-Lip group. i Apparent permeability coefficient (Papp) values of vesicles across Caco-2/HT29 cell monolayers in Transwell inserts. j IVIS images 4 h post-oral administration of free DiR or DiR-loaded vesicles in intestine. SI-P: proximal small intestine; SI-D: distal small intestine; LI: large intestine. Scale bars, 2.5 cm. k–m, Absorption dynamics for various vesicles. IVIS image quantification at 1, 4, 8, and 12 h post-oral administration of DiR-loaded vesicles in SI-P (k), SI-D (l), or LI (m). n Schematic representation showing that mEV-Lip, with PEG modification, penetrating the mucus layer and being taken up and transported by intestinal epithelial cells (IECs) via FcRn/PEPT1-mediated transport after oral administration. Data are mean ± SD (n = 3). One-way ANOVA with a Tukey post hoc test was used for statistical significance. * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001. NS, not significant. Source data are provided as a Source Data file.

Inspired by this insight, we engineered mEV-Lip hybrid vesicles, leveraging the transporter-binding ability to target the proximal small intestine. Initially, mEV was isolated using a classical ultracentrifugation method. To remove casein from milk, ethylenediaminetetraacetic acid disodium salt (EDTA-2Na, 30 mM) was added to chelate calcium ions and disrupt casein micelle integrity, removing casein through micellar disassembly before ultracentrifugation. Dynamic light scattering (Fig. 2b) and nanoparticle-tracking analysis (NTA, Fig. 2c) revealed hydrated particle sizes of 132.9 ± 1.8 nm and 113.5 ± 2.5 nm, respectively, aligning with the typical size range of extracellular vesicles (30–150 nm). The modest polydispersity index (PDI, 0.208 ± 0.016) and substantial yield (1.08 × 10¹¹ ± 2.62 × 10⁹ particles/mL) indicated successful isolation of mEV.

Next, we modified liposomes with PEG-functionalized phospholipids utilizing the classical thin-film hydration method. The hybrid vesicles were subsequently fabricated by capitalizing on the flow and fusion characteristics inherent to the mEV membrane’s phospholipid bilayer. Fluorescence scanning patterns of mEV, liposomes (Lip), and mEV-Lip demonstrated the fluorescence resonance energy transfer (FRET) phenomenon in the hybrid vesicles (Fig. 2d). Confocal laser scanning microscope (CLSM) further confirmed the successful hybridization of the mEV-Lip (Supplementary Fig. 2). Cryo-transmission electron microscope (cryo-TEM) images showed that mEV-Lip had spheroidal morphology (Supplementary Fig. 3). Western Blot (WB) analysis corroborate the presence of mEV signature proteins (CD9, CD63, HSP70) in both mEV and mEV-Lip, while calnexin, an endoplasmic reticulum marker, was absent in both groups except the supernatant (Fig. 2e). Particle size and zeta potential measurements indicated that after fusion with Lip, mEV had an increased particle size and a more negative potential (Supplementary Fig. 4). Moreover, stability tests showed that both mEV and mEV-Lip maintained favorable colloidal stability in simulated gastric fluid (SGF) and simulated intestinal fluid (SIF), as evidenced by minimal changes in particle size (Supplementary Fig. 5). These vesicles also exhibited excellent stability over 7 days at 4 °C and −80 °C (Supplementary Fig. 6). In summary, we successfully prepared hybrid vesicles modified with PEG-functionalized phospholipids, which possess stable colloidal properties.

One of the primary objectives of this study was to enhance the mucus penetration ability of natural mEV by introducing hydrophilic PEG. To closely mimic the in vivo environment, freshly obtained porcine small intestinal mucus was used to evaluate mucus penetration. NTA showed that the mobility of mEV was restricted in the mucus, whereas Lip and mEV-Lip had significantly wider mobility ranges (Fig. 2f). Mean square displacement (MSD) calculations revealed that within a 1 s interval, the MSD values of Lip and mEV-Lip were greater than that of mEV (Fig. 2g). Besides, a transwell permeable scaffold was used to construct a three-dimensional mucus model (Supplementary Fig. 7a), and the apparent permeability coefficient (Papp) was measured to quantify the transport rate from the donor chamber to the recipient chamber²⁴. The hydrophilic modification by PEG significantly enhanced the mucus penetration ability of natural extracellular vesicles, with a 1.6-fold increase (Supplementary Fig. 7b). These findings imply that the PEG layer is instrumental in improving permeability of vesicles in the intestinal mucus.

After penetrating the mucus layer, intestinal epithelial cells pose another barrier to absorption. We investigated the intestinal epithelial cell barrier using Caco-2/HT29 coculture model. Cellular uptake studies showed that mEV-Lip had a significant increase compared to Lip (2.7-fold) and mEV (1.7-fold), indicating that the combination of PEG with mEV is an effective strategy to overcome both the mucus and apical membrane barriers (Supplementary Fig. 8). Additionally, mEV-Lip demonstrated enhanced exocytosis ability compared to Lip and mEV (Fig. 2h). A cell monolayer was established to assess the transcytosis efficiency of various formulations across the intestinal epithelium. The Papp results indicated that mEV-Lip improved the transcytosis efficiency of Lip and mEV in the Caco-2/HT29 coculture cell monolayer model (Fig. 2i). Bidirectional exocytosis profiling revealed that the exocytosis rate of Lip from basolateral side was only 63% of that from the apical side, whereas mEV-Lip raised this fraction to 131% (Supplementary Fig. 9), yielding a 2-fold higher basolateral-to-apical exocytosis ratio. To determine whether vesicles remain structurally intact during transcytosis, double fluorophores-labeled mEV-Lip were applied to the cell monolayer. A robust FRET signal detected in the basolateral compartment (Supplementary Fig. 10) confirmed that mEV-Lip cross the epithelium with intact structure. These findings suggest that mEV-Lip can be more efficiently transported into and out of the epithelial cells, with an “easy entry, easy exit” characteristic.

To visualize the in vivo distribution of vesicles, we encapsulated a fluorescent dye (1,1-dioctadecyl-3,3,3,3-tetramethylindotricarbocyanine iodide, DiR) into the vesicles. Mice were orally given DiR-labeled vesicles, and their intestines were subsequently imaged using an in vivo imaging system (IVIS). Compared with free DiR, mEV-Lip had a higher signal intensity in the small intestine and a markedly lower intensity in the large intestine (Fig. 2j–m, and Supplementary Fig. 11). In vitro intestinal absorption experiments also showed that mEV-Lip had significantly higher absorption in the small intestine compared with Lip (Supplementary Fig. 12). We further investigated the micro-distribution of the vesicles within the proximal small intestine. The findings indicated that the uptake of mEV-Lip by intestinal epithelium surpassed that of other vesicles (Supplementary Fig. 13). Immunofluorescent staining of intestinal sections showed that mEV-Lip co-localized with FcRn and PEPT1 in small intestinal epithelial cells (Supplementary Fig. 14). Moreover, linear regression analysis demonstrated that epithelial uptake of mEV-Lip correlated tightly with FcRn and PEPT1 expression levels (Supplementary Fig. 15). The above results collectively suggest that the absorption of mEV-Lip in the small intestine is potentially mediated by specific interactions with FcRn/PEPT1.

To further demonstrate this, we employed the FcRn inhibitor human serum albumin (HSA) and the PEPT1 inhibitor glycylsarcosine (Gly-Sar) to block each pathway in vivo, which resulted in a marked reduction in the absorption of mEV-Lip in the small intestine (Supplementary Fig. 16). In contrast, the absorption of Lip remained unchanged upon treatment with FcRn or PEPT1 inhibitor. Moreover, when FcRn and PEPT1 were overexpressed in Caco-2 cells (Supplementary Fig. 17), HSA and Gly-Sar significantly inhibited the endocytosis of mEV and mEV-Lip, while having a relatively weaker inhibitory effect on Lip (Supplementary Fig. 18). The transcytosis assays revealed similar pathway dependency, with mEV and mEV-Lip transport across epithelial barriers being significantly compromised by both inhibitors (Supplementary Fig. 19). Collectively, these results suggest that the hybrid vesicles promote their absorption in the proximal small intestine through the transporters FcRn and PEPT1 (Fig. 2n), which holds great promise for oral antibiotic delivery.

Hybrid vesicles reduce intestinal residues and enhance oral bioavailability

To evaluate whether the increased absorption of mEV-Lip at the proximal small intestine could decrease intestinal residues, the quantity of residual DiR-labeled mEV-Lip in feces was measured using the IVIS imaging system at various time points post oral administration. Comparative analysis revealed that mice treated with mEV-Lip showed a marked reduction in fecal residues compared to free DiR and DiR-labeled Lip (Fig. 3a, b), which may be correlated with enhanced proximal intestinal absorption.

Fig. 3. Hybrid vesicles mitigate intestinal residues and enhance oral antibiotics absorption.

a Representative images of residual vesicles in murine fecal samples at different time points. b Quantification of vesicle fluorescence intensity in fecal matter. n = 3 mice. c Drug release profile in 2 h simulated gastric fluid (SGF) followed by 6 h simulated intestinal fluid (SIF). n = 3. d Plasma concentration-time profiles of free Cef and Cef-loaded vesicles in mice. n = 5 mice. e Relative bioavailability calculation by comparing the area under the curve (AUC) of the vesicle formulation to that of free Cef. All data presented as mean ± SD. Statistical significance determined via one-way ANOVA with a Tukey post hoc test. *P < 0.05, **P < 0.01. Source data are provided as a Source Data file.

We subsequently explored whether mEV-Lip could serve as carriers to boost the bioavailability of encapsulated antibiotics. For this purpose, Cef was employed as a model antibiotic, and the corresponding Cef-loaded vesicles were designated as Lip@Cef, mEV@Cef, and mEV-Lip@Cef. Although the encapsulation efficiency of natural mEV was only 11.29%, mEV-Lip significantly increased the drug encapsulation efficiency to 56.02% (4.96-fold, P = 0.0002) (Supplementary Table. 1). The same trend was observed for seven additional antibiotics representing distinct chemical classes: amoxicillin (Amo), cefalexin (Cax), rifampicin (Rif), ciprofloxacin hydrochloride (Cfh), ceftriaxone sodium (Cxs), ampicillin sodium (Aps), sulfamethoxazole (Sul) (Supplementary Table. 2). This substantial improvement highlights the advantage of preparing hybrid vesicles, where the enhanced membrane fluidity imparted by liposomal components facilitates superior drug-loading capacity compared to natural extracellular vesicles. Besides, in vitro drug release experiments revealed that Cef was released in a slow and sustained manner from mEV-Lip@Cef when incubated in simulated buffers (Fig. 3c). This sustained release profile overcomes the problem of rapid drug release from natural mEV. Furthermore, Lip@Cef and mEV-Lip@Cef had a certain sustained release effect and played important roles in blood circulation (Supplementary Fig. 20). The serum concentration of Cef in the mEV-Lip@Cef group was significantly higher than that of free Cef (Fig. 3d and Supplementary Table. 3). Additionally, the bioavailability of mEV-Lip@Cef was 3.24 times higher than that of free Cef (Fig. 3e), indicating that mEV-Lip@Cef significantly promoted the oral absorption of Cef. These results collectively suggest that the mEV-Lip packaging enhances oral antibiotics bioavailability and reduces intestinal residues, which underscores the potential of mEV-Lip as an effective oral delivery system for antibiotics.

mEV-Lip@Cef effectively eliminates Klebsiella pneumoniae infection in the lung

To evaluate the therapeutic potential of hybrid vesicles, we established a murine model of pulmonary infection induced by Klebsiella pneumoniae, a predominant nosocomial pathogen causing pneumonia (Fig. 4a)²⁵. Whereas only 50% of mice receiving free Cef survived the 4 d observation period, 88% of animals treated with mEV-Lip@Cef remained alive (Supplementary Fig. 21). Compared to free Cef, which achieved modest bacterial reduction in the lung, both Lip@Cef and mEV@Cef demonstrated superior antibiotic activity (Fig. 4b–d). Remarkably, mEV-Lip@Cef exhibited the highest therapeutic efficacy, achieving the lowest bacterial load in the lung. Besides, the measurement of inflammatory cytokines in lungs revealed substantially reduced inflammation in the Cef-loaded vesicles groups compared with mice treated with free Cef (Fig. 4e). Pulmonary pathological and histological analyses further showed that both free Cef and other Cef-loaded vesicle groups exhibited diminished inflammation and pathology compared with the phosphate buffer solution (PBS) group (Fig. 4f–h). Importantly, lower inflammation and attenuated tissue damage were observed in mEV-Lip@Cef-treated mice compared with free Cef-treated and other Cef-loaded vesicle-treated mice. In summary, these results suggest that mEV-Lip@Cef exhibits enhanced efficacy in treating bacterial pneumonia.

Fig. 4. Cef-loaded hybrid vesicles effectively eliminate Klebsiella pneumoniae infection in the lung.

a Schematic illustration of the experimental protocol for therapy in the mouse model of Klebsiella pneumoniae infection. b Representative Klebsiella pneumoniae colonies formed on plates from the lung homogenates of mice with pneumonia after different treatments. c Quantitative analysis of bacterial loads in the lungs. ****P = 0.00000021 for the comparison between the PBS group and free Cef group. ****P = 0.000000033 for the comparison between the Cef group and Lip@Cef group. d Representative images of lung tissue from different groups after Gram staining. Scale bars, 100 μm. Arrows indicate Klebsiella pneumoniae. A representative image from one of five independent fields of view in a single experiment. e Levels of inflammatory cytokines in the lung tissue. ****P = 0.000025 for the comparison between the PBS group and free Cef group (IL-6). ****P = 0.0000055 for the comparison between the Cef group and Lip@Cef group (IL-6). ****P = 0.0000028 for the comparison between the PBS group and free Cef group (TNF-α). ****P = 0.000077 for the comparison between the Cef group and Lip@Cef group (TNF-α). ****P = 0.000031 for the comparison between the mEV@Cef group and mEV-Lip@Cef group (TNF-α). f The appearance of lungs from different treatment groups. g Hematoxylin and eosin (H&E) staining of lung sections. Scale bars, 100 μm. A representative image from one of five independent fields of view in a single experiment. h Lung index (lung/body weight ratio) quantification. **** P = 0.000021 for the comparison between the PBS group and free Cef group. Data are presented as the mean ± SD. n = 5 mice. Statistical significance was determined using one-way ANOVA with a Tukey post hoc test. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Source data are provided as a Source Data file.

mEV-Lip@Cef effectively eradicates Klebsiella pneumoniae infection in a bacteremia model

To further evaluate the antibacterial efficacy of antibiotic-loaded vesicles, we employed a bacteremia model (Fig. 5a). Mice were intravenously inoculated with a standardized dose of Klebsiella pneumoniae. Subsequently, at pre-determined time points, the mice were treated with free Cef and Cef-loaded vesicles. Only 12.5% of mice receiving free Cef survived the 8 d monitoring period, while 62.5% of animals administered mEV-Lip@Cef remained alive (Supplementary Fig. 22). The appearance of liver and spleen revealed a notable reduction in inflammation in free Cef and Cef-loaded vesicle groups compared to PBS group (Fig. 5b). Remarkably, the mEV-Lip@Cef group exhibited a significant attenuation of inflammation. To quantify the pathogen burden, we measured the colony-forming units (CFU) in the peripheral blood, liver, and spleen 24 h post-infection. Consistent with our findings in the lung infection model, mice treated with mEV-Lip@Cef demonstrated a more efficient clearance of Klebsiella pneumoniae compared to those treated with free Cef, Lip@Cef, and mEV@Cef (Fig. 5c–e). In addition, hematoxylin and eosin (H&E) staining was performed on liver and spleen tissue sections (Fig. 5f). The staining results showed that mEV-Lip@Cef-treated mice had less inflammation and reduced tissue damage in the liver and spleen compared to mice treated with free Cef and other Cef-loaded vesicles. Collectively, these results suggest that mEV-Lip@Cef possesses superior efficacy in eliminating bacterial pathogens, highlighting its potential as an effective approach for antibiotic therapy.

Fig. 5. Cef-loaded hybrid vesicles effectively eradicate Klebsiella pneumoniae infection in a bacteremia model.

a Experimental timeline schematic of the bacteremia infection model. b Macroscopic pathological changes observed in hepatic and splenic tissues. Comparative analysis of microbial burden in circulatory (c), hepatic (d), and splenic (e). f Histopathological evaluation of liver and spleen sections through H&E staining. Scale bars, 100 μm. A representative image from one of five independent fields of view in a single experiment. Data presented mean ± SD. n = 5 mice. Statistical comparisons were performed using one-way ANOVA with a Tukey post hoc test. *P < 0.05, **P < 0.01. Source data are provided as a Source Data file.

Hybrid vesicles mitigate antibiotic-driven dysbiosis

The enhanced proximal small intestine absorption profile of oral hybrid vesicles may confer ecological advantages by minimizing gut microbiota disruption. Notably, pharmacological alteration of enteric flora through conventional antibiotic formulations can induce substantial community restricting, affecting approximately 30% of bacterial taxa aboundance²⁶. This dysbiotic transformation is manifested through abrupt declines in α-diversity metrics including species richness and evenness, with incomplete microbiome restoration post-treatment potentially potentiating susceptibility to dysbiosis-associated pathologies, particularly in pediatric populations^27–29. To evaluate these differential effects, we conducted longitudinal comparative analyses of gut microbiota ecology following administration of free Cef versus hybrid vesicle-encapsulated Cef formulations at multiple intervals post-administration (Fig. 6a).

Fig. 6. Protective effects of hybrid vesicle delivery against antibiotic-induced microbial imbalance.

a Schematic design workflow. b Temporal changes in gut microbiota α-diversity indices across different groups. The pink band indicates therapeutic intervention periods with either free Cef or Cef-loaded vesicles. c Comparative analysis of α-diversity through AUC. ****P = 0.000012 for the comparison between the control group and free Cef group. d Longitudinal variations in fecal microbial β-diversity between modalities, assessed by Unweighted UniFrac metric relative to baseline. e Cumulative β-diversity alterations quantified by AUC. f Multivariate separation of intestinal bacteria profiles visualized through principal components analysis (PCA). The circles represent the confidence ellipses, illustrating the distribution of samples within each group. g Taxonomic composition profiling of gut microbiota. h Hierarchical clustering analysis of microbial abundance patterns. Data represent mean ± SD. n = 3 mice. Statistical comparisons performed using one-way ANOVA with a Tukey post hoc test. **P < 0.01, ***P < 0.001, ****P < 0.0001. Source data are provided as a Source Data file.

As anticipated, 16S ribosomal RNA sequencing revealed marked depletion of stool α-diversity and substantial dysbiosis in fecal communities following free Cef administration (Fig. 6b–e). Strikingly, the mEV-Lip@Cef demonstrated superior preservation of gut microbiota homeostasis, facilitating accelerated recovery of both microbial richness and ecological architecture compared to free Cef. The protective effect was further corroborated by principal components analysis (PCA) of microbial communities, which showed tighter clustering patterns in the mEV-Lip@Cef group compared to the pronounced dispersion observed with free Cef (Fig. 6f). Taxonomic profiling at the phylum level identified significant perturbations induced by free Cef, characterized by decrease of Bacteroidota and Bacillota populations alongside increase of Pseudomonadota (Fig. 6g, h). In contrast, mEV-Lip@Cef administration maintained near-physiological phylum distributions. Taken together, these findings suggest that the engineered hybrid vesicle system effectively mitigates antibiotic-associated microbiota perturbations, preserving critical ecological parameters including bacterial richness, diversity, and composition.

Hybrid vesicle-mediated antibiotic delivery ameliorates dysbiosis-associated obesity

Emerging evidence underscores the critical involvement of intestinal microbial communities in regulating systemic metabolic process, particularly energy balance regulation and lipid accumulation³⁰. Pathological conditions including obesity, diabetes mellitus type 2, and hepatic steatosis demonstrate enhanced dietary energy harvest efficiency, a characteristic feature of metabolic dysregulation³¹. To investigate the potential impact of antibiotic-loaded vesicles on antibiotic-associated metabolic perturbations, we established a dietary obesity model (Fig. 7a).

Fig. 7. Hybrid vesicle-mediated delivery of antibiotics ameliorates dysbiosis-associated obesity.

a Graphical overview of the experimental design. b Body weight changes across intervention group. c Intraperitoneal glucose tolerance test (IPGTT) results for mice. d AUC analysis of blood glucose levels during IPGTT. e Hepatic cholesterol accumulation after prolonged high-fat diet (HFD) exposure. f Representative H&E-stained sections of liver tissue. Scale bars: 100 μm (overview), 50 μm (magnified regions indicated by dashed squares). A representative image from one of five independent fields of view in a single experiment. g Schematic showing the experimental protocol. h Body weight changes in recipient mice. i IPGTT results for recipient mice. j AUC analysis of blood glucose during IPGTT in recipient mice. k Hepatic cholesterol deposition. l Parenchymal morphology assessment in recipient mice. Scale bars, 100 μm (overview), 50 μm (magnified regions indicated by dashed squares). A representative image from one of five independent fields of view in a single experiment. Data are expressed as the mean ± SD. n = 5 mice. Statistical comparisons employed one-way ANOVA with a Tukey post hoc test. *P < 0.05, **P < 0.01, ***P < 0.001. NS not significant. Source data are provided as a Source Data file.

Treatment with free Cef resulted in a significant increase in high-fat diet (HFD)-induced body mass accumulation (Fig. 7b). In contrast, the mEV-Lip@Cef showed moderated weight alterations, suggesting a reduced impact on weight gain. Similarly, metabolic parameter analysis demonstrated that free Cef treatment exacerbated HFD-induced metabolic dysregulation, manifesting as impaired glucose tolerance and hepatic cholesterol accumulation (Fig. 7c–e). In comparison, the engineered nanocarrier system maintained near-physiological glycemic responses and effectively constrained hepatic cholesterol levels. Furthermore, histological analysis revealed milder liver steatosis in mice treated with mEV-Lip@Cef compared to those treated with free Cef (Fig. 7f).

To investigate whether these metabolic alterations were driven by changes in gut microbiota, we conducted fecal microbiota transplantation (FMT) studies (Fig. 7g). Three donor cohorts were established: free Cef treatment, mEV-Lip@Cef treatment, and antibiotic-naïve controls. Recipient mice received fecal transplants from donor groups and were subsequently placed on an HFD regimen. Notably, animals receiving microbiota from free Cef-treated donors developed exacerbated weight gain, impaired glucose resistance, elevated hepatic cholesterol accumulation, and pronounced steatotic changes compared to those that received microbiota from mEV-Lip@Cef-treated donors (Fig. 7h–l). These findings jointly indicate that engineered vesicle-mediated antibiotic delivery preserves metabolic homeostasis by maintaining microbial ecological balance.

Hybrid vesicle delivery of antibiotics decreases dysbiosis-associated Escherichia coli infection

Antibiotic-mediated disruption of gut microbiota homeostasis has been widely documented to enhance vulnerability to enteric infections, primarily through two mechanisms: facilitating colonization by exogenous pathogens or promoting proliferation of resident opportunistic pathobionts^32–34. To evaluate the effects of pretreatment with free Cef and mEV-Lip@Cef on the severity of Escherichia coli intestinal infection, we conducted an experimental study (Fig. 8a). Mice were pretreated with antibiotics to simulate the initial treatment and associated perturbation of the gut microbiota, followed by an intestinal challenge with pathogenic Escherichia coli at day 5. Quantitative analysis revealed that free Cef-pretreated mice exhibited significantly compromised resistance to Escherichia coli colonization compared to antibiotic-naïve controls (Fig. 8b–e). This heightened susceptibility manifested as elevated pathogen burdens across fecal samples, cecal contents, and colonic tissues, accompanied by pronounced mucosal hyperplasia. Notably, mice receiving mEV-Lip@Cef pretreatment demonstrated infection resistance comparable to untreated controls, with no statistically significant differences in pathogen colonization levels and histopathological manifestations. These findings collectively suggest that engineered vesicle-based antibiotic delivery effectively mitigates dysbiosis-associated pathogen infections (Fig. 8f).

Fig. 8. The hybrid vesicle-mediated delivery of antibiotics decreases dysbiosis-associated infection by Escherichia coli.

a Experimental design workflow. Microbial burden assessment in feces (b), caecal contents (c), and colon (d). e Histopathological characterization of colon sections by H&E staining. Scale bars, 100 μm. A representative image from one of five independent fields of view in a single experiment. f, Schematic diagram illustrating the mechanism by which treatment with free Cef decreases microbiota diversity and leads to the expansion of the Escherichia coli population. In contrast, treatment with mEV-Lip@Cef has minimal impact on microbiota, maintaining diversity and resistance to pathogens colonization. Data expressed as mean ± SD. n = 5 mice. Statistical analyses employed one-way ANOVA with a Tukey post hoc test. * P < 0.05. NS, not significant. Source data are provided as a Source Data file.

Discussion

Antibiotics have long been hailed as life-saving agents in the battle against bacterial infections. However, the shadow of their overuse looms large, casting a multitude of risks upon human health. In response to this pressing issue, we have developed an innovative antibiotic delivery technology that enhances oral bioavailability while minimizing the impact on the gut microbiota. Our study demonstrates the potential of mEV-Lip hybrid vesicles to achieve targeted and efficient absorption of antibiotics at the proximal small intestine.

The mEV-Lip system significantly enhances the oral absorption of antibiotics, as evidenced by our experiments with Cef. This hybrid vesicle formulation not only improves drug bioavailability but also reduces perturbations to the gut microbiota, thereby mitigating dysbiosis-induced adverse effects. Our findings, although preliminary and based on murine models, suggest that mEV-Lip-based antibiotic delivery can substantially reduce the severity of antibiotic-induced perturbation to gut microbiota. This is particularly significant given the well-established clinical manifestations of dysbiosis, which can severely endanger patients in hospital settings.

The ability to deliver antibiotics orally while minimizing the need for intravenous administration could revolutionize the treatment of bacterial infections. Oral delivery of antibiotic-loaded mEV-Lip could potentially reduce the duration and necessity of hospitalizations, thereby lowering associated costs and risks. To facilitate penetration of the intestinal mucus layer, we incorporated the PEG corona. While repeated dosing may induce the production of PEG-specific antibodies, which potentially cause a higher incidence of side effects and reduce the therapeutic efficacy^35–37. Our study has shown that mEV-Lip performs efficiently for the delivery of Cef, but the potential applications extend beyond this single antibiotic. Future research should explore the efficacy of this system with other antibiotics and even other drugs that face absorption-related or gut-dysbiosis-related limitations. The strong and selective absorption efficiency enabled by mEV-Lip suggests its potential use for the oral delivery of a wide range of drugs. This technology offers a potential strategy to mitigate the challenges of low oral bioavailability and drug-induced dysbiosis.

Methods

Materials

Egg lecithin was procured from Shanghai Taiwei Pharmaceutical Co., Ltd. (Shanghai, China). Cholesterol, Rhodamine B (RhB), and glycylsarcosine (Gly-Sar) were obtained from Shanghai Maclean Biochemical Technology Co., Ltd. (Shanghai, China). DSPE-mPEG₂₀₀₀ was purchased from AVT (Shanghai) Pharmaceutical Tech Co., Ltd. (Shanghai, China). Recombinant human serum albumin (HSA) was purchased from Beijing Solarbio Science & Technology Co., Ltd. (Beijing, China). Cefdinir (Cef) and coumarin 6 (C6) were obtained from Shanghai Yuanye Biotechnology Co., Ltd. (Shanghai, China). 1,1-dioctadecyl-3,3,3,3-tetramethylindotricarbocyanine iodide (DiR) was acquired from Dalian Meilun Biotechnology Co., Ltd. (Dalian, China). Enhanced BCA protein assay kit, 4’,6-diamidino-2-phenylindole (DAPI), and goat anti-rabbit IgG Alexa Fluor 647 were purchased from Beyotime Biotechnology Co., Ltd. (Shanghai, China). Anti-CD9 antibody, anti-CD63 antibody, anti-HSP70 antibody, and anti-calnexin antibody were purchased from Bioss Biotechnology (Beijing, China). Anti-FCGRT/FcRn antibody was obtained from ABclonal Technology Co., Ltd. (Wuhan, China). Anti-SLC15A1/PEPT1 antibody, β-actin antibody, and goat anti-rabbit IgG H&L (HRP) were obtained from Proteintech Group (Wuhan, China). Interleukin-6 (IL-6) and tumor necrosis factor-alpha (TNF-α) enzyme-linked immunosorbent assay (ELISA) kits were supplied by Shanghai Jianglai Biotechnology Co., Ltd. (Shanghai, China). Total cholesterol detection kit was purchased from Nanjing Jiancheng Bioengineering Institute (Nanjing, China). Methanol and acetonitrile, both HPLC grade, were purchased from Aladdin Biotechnology Co., Ltd. (Shanghai, China). All other chemicals and solvents used for this study were of analytical grade.

Strains

Klebsiella pneumoniae (KP2237, serotype K173, resistance gene: aac(3)-IIa, aph(3”)-Ib, aph(6)-Id, blaCTX-M-15, blaKPC-2, blaSHV-106, blaTEM-1B, dfrA14, fosA, oqxA, oqxB, qnrB1, sul2) used in this study is a clinical strain isolated from human, preserved in our laboratory; Escherichia coli (ATCC25922) used in this study was obtained from the American Type Culture Collection and is preserved in our laboratory.

Cells and animals

Human intestinal cell lines, Caco-2 and HT29 cells, were procured from Beijing Dingguo Biotechnology Co., Ltd. (Beijing, China). Cellular maintenance was conducted in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum (FBS), 1% penicillin, 1% streptomycin, and 1% of non-essential amino acids. All cultures were maintained under standardized conditions (37 °C, 5% CO₂, 95% relative humidity).

Male Balb/c mice certified by Hunan SJA Laboratory Animal Co., Ltd. (SCXK Xiang 2021-0002) were housed at the School of Pharmacy, Southwest University, under the experimental animal use license number SYXK (Yu) 2020-0006. Animals were housed under a 12 h light/12 h dark cycle, with the ambient temperature at approximately 25 °C and relative humidity at 50 ± 10%. All procedures complied with guidelines and received institutional approval (IACUC-20250407-01) from the Laboratory Animal Use and Management Committee of Southwest University.

Isolation of milk extracellular vesicles (mEV)

Pasteurized bovine milk was sourced from retail sources. mEV isolation employed a multi-step differential centrifugation protocol: primary clarification at 5000 g (30 min, 4 °C) removed somatic cells and debris, followed by casein dissociation using ethylenediaminetetraacetic acid disodium salt (EDTA-2Na, 30 mM). The supernatant underwent filtration (0.22 μm polyethersulfone membrane) prior to sequential ultracentrifugation (100,000 g for 60 min, 210,000 g for 60 min, Optima L-90K, Beckman, USA) for vesicle subfractionation. Pelleted vesicles were suspended in phosphate-buffered saline (PBS) and quantified via a BCA protein assay kit. mEV samples with a total protein concentration below 2 mg/mL were cryopreserved at −80 °C until further use.

Preparation of hybrid vesicles

The hybrid vesicles were fabricated using the dried film hydration-extrusion technique. Primary constituents including phospholipids (10 mg), cholesterol (2 mg), and DSPE-mPEG₂₀₀₀ (1.67 mg) were initially solubilized in chloroform to form the liposome component. Following complete solvent removal under vacuum conditions, the resultant dry lipid films were hydrated with an mEV suspension (4 mg of mEV diluted to 4 mL with PBS). Subsequently, the mixture was extruded through a liposome extruder fitted with a 200 nm polycarbonate membrane filter. For the preparation of Cef, C6, RhB or DiR-loaded hybrid vesicles, the following procedure was employed. A mixture of mEV and the respective drug solution (Cef: 5 mg, C6: 0.1 mg, RhB: 0.4 mg, or DiR: 0.4 mg) was sonicated (120 W, 1 min, with a 2 s on/2 s off cycle). This sonicate mixture was then processed using the same hydration and extrusion steps as described for the unloaded hybrid vesicles.

Characterization of vesicles

Colloidal properties of the engineered vesicles were assessed using dynamic light scattering (3000+Ultra, Malvern Instruments Ltd., UK). The particle concentration of mEV suspensions was performed via nanoparticle-tracking analysis (NTA, Nanosight LM10, Malvern Instruments Ltd., UK). For morphology characterization, vesicles were observed by cryo-TEM (TF20, FEI, USA) with an acceleration voltage of 200 kV. To confirm the successful extraction of mEV, Western Blot (WB) analysis was conducted on both mEV and the hybrid vesicles. The presence of several mEV markers, including CD9, CD63, and HSP70, was detected, along with the supernatant protein calnexin, following standard protocols.

To validate the fusion of mEV with liposomes, a fluorescence resonance energy transfer (FRET)-based analytical approach was employed. Specifically, mEV was labeled by incubation with a RhB solution, while liposomes were labeled by incorporating a chloroform solution of C6 during the film preparation process. Subsequently, RhB/C6 co-labeled hybrid vesicles were fabricated using the dried film hydration-extrusion method. Spectral characterization was performed using a fluorescence spectrophotometer with 445 nm excitation, capturing emission profiles across a 475–700 nm wavelength range to quantify energy transfer dynamics.

The encapsulation efficiency was evaluated by eliminating unencapsulated drugs through 100 kDa molecular weight cutoff (MWCO) ultrafiltration membranes. Quantitative analysis of Cef was performed using an optimized high performance liquid chromatography (HPLC, LC-20AD, Shimadzu, Japan) protocol utilizing C18 column (150 mm × 4.6 mm, 5 μm). Chromatographic separation was achieved with a mobile phase consisting 0.25% tetramethylammonium hydroxide (pH-adjusted to 5.5 using phosphoric acid), acetonitrile, and methanol (900:60:40, v/v/v), supplemented with 0.4 mL/L of 0.1 mol/L EDTA-2Na solution. Operational parameters include isocratic elution at 1.0 mL/min, UV detection at 254 nm, 20 μL injection volume, and column thermostating at 25 °C. Encapsulation efficiency was derived from the percentage ratio between the amount of Cef detected in the vesicles to the total amount of Cef used in the preparation.

Stability and in vitro release profile

The colloidal stability of the various vesicles was evaluated under physiologically relevant conditions, including simulated gastric fluid (SGF) and simulated intestinal fluid (SIF). Specifically, vesicle suspensions (0.1 mL) were homogenously dispersed in 0.9 mL of either SGF or SIF and incubated. The particle size of the vesicles was measured at predetermined time intervals.

The in vitro release profiles of Cef-loaded vesicles were evaluated in SGF and SIF. A volume of 2 mL of Cef-loaded vesicles was placed in a dialysis bag with a MWCO of 10 kDa. The sealed bag underwent phased incubation: primary SGF (50 mL, 37 °C, 2 h) followed by SIF (50 mL, 37 °C, 6 h) under continuous agitation (90 rpm). At each sampling time point, aliquots (0.5 mL) were periodically collected and immediately replaced with thermo-equilibrated fresh medium. The amount of Cef released was subsequently quantified using HPLC analysis.

Mucus permeability evaluation

The mobility of the vesicles within mucus was assessed using NTA. A volume of 0.1 mL vesicles was introduced into 0.9 mL porcine intestinal mucus and incubated for 30 min. The movement of the vesicles within the mucus was then recorded. The averaged mean square displacement (MSD) of the particles was calculated based on their trajectories³⁸.

The mucus permeability of the vesicles was quantitatively assessed using a Transwell system. Initially, the donor chamber received 0.1 mL of porcine intestinal mucus, while the acceptor compartment was filled with 0.8 mL of PBS. To achieve uniform mucus layer formulation, the assembly underwent equilibration at 37 °C with shaking for 30 min. Subsequently, 0.15 mL of C6-labeled vesicles was carefully administered to the mucus-containing donor chamber, followed by continuous incubation at 37 °C. Aliquots (80 μL) were periodically harvested from the acceptor compartment at designed intervals, with immediate replenishment of preheated PBS. The vesicles that traversed the mucus layer were quantified using a microplate reader (Synergy H1, American Berton Instrument Company, USA). The apparent permeability coefficient (Papp) values were calculated.

Cellular uptake and endocytic mechanisms

Cellular internalization studies of diverse vesicles were performed using the Caco-2/HT29 coculture model. The intestinal epithelial models were co-seeded in 96-well plates at a 7:3 ratio (Caco-2:HT29) and maintained under standard culture condition for 48 h. Cellular confluence was quantitatively assessed through Alamar Blue assay. Subsequently, the culture medium was exchanged with serum-free medium containing C6-labeled vesicles (C6 concentration: 0.5 μg/mL). After continuous incubation for 3 h, the cells were lysed with dimethyl sulfoxide (DMSO), and intracellular fluorescence accumulation was determined using a microplate reader.

Endocytic pathway analysis was conducted employing the Caco-2 cell model. Following 48 h period in 96-well plates, cellular density was assessed using the Alamar Blue assay. To investigate the specific pathways of endocytosis, HSA (40 μg/mL) and Gly-Sar (100 mM) were added to separate wells and incubated for 1 h. Thereafter, C6-labeled vesicles (C6 concentration: 0.5 μg/mL) were added and incubated for an additional 3 h. The cells were then lysed with DMSO, followed by fluorometric analysis of intracellular accumulation using a microplate reader.

Exocytosis studies

Exocytosis experiments of various vesicles were conducted using the Caco-2/HT29 coculture model (7:3). Following 48 h culture in 96-well plates, cellular density was quantified through Alamar Blue assay. The cells were then exposed to C6-labeled vesicles (C6 concentration: 0.5 μg/mL) for 3 h, washed three times with cold PBS, and incubated with DMEM for 1 h. After lysis with DMSO, the fluorescence intensity was quantified on a microplate reader.

Transcellular transport studies

A co-culture model of intestinal epithelial cells was established using Caco-2 and HT29 cells at a 7:3 ratio in Transwell inserts. Following 21 days of differentiation under standard culture condition, the transepithelial electrical resistance (TEER) of approximately 300 Ω cm² was achieved. Prior to transport experiments, the differentiated monolayers were pre-incubated with blank medium for 30 min at 37 °C. The apical compartment then received 0.2 mL of medium containing C6-labeled vesicles (C6 concentration: 1.0 μg/mL), while the basolateral chamber contained 0.2 mL fresh medium. Aliquots (80 μL) were collected from the basolateral side at designated time intervals with immediate volume replacement using pre-warmed medium. The collected samples were subjected to disruption using DMSO, followed by fluorescence quantification employing a microplate reader. The Papp values were calculated.

Biodistribution of vesicles

Balb/c mice (8 weeks old) were fasted overnight. Subsequently, they were orally administered DiR-labeled vesicles (DiR: 0.5 mg/kg). At predetermined time points, the mice were sacrificed, and their digestive tract and fecal samples were collected. The distribution of DiR fluorescence in the intestines and fecal samples was collected and determined using an in vivo imaging system (IVIS, Vilber Lourmat).

Intestinal absorption of Cef-loaded vesicles

After fasting the Balb/c mice (8 weeks old) overnight, they were anesthetized and the segments of the proximal small intestine, distal small intestine, and colon were carefully extracted. Each intestinal segment was washed thoroughly, and one end was ligated. At the other end, 0.1 mL of C6-labeled vesicles (C6 concentration: 1.0 μg/mL) was injected into each segment. The intestinal segments were then immersed in fresh Krebs-Ringer (K-R) solution at 37 °C in a water bath, with a continuous supply of 95% O₂ to maintain tissue viability. At predetermined time intervals, 0.1 mL of the extracellular fluid was collected from the K-R solution, and an equal volume of fresh extracellular fluid was immediately replenished. The collected samples were mixed with DMSO to disrupt the vesicles, and the fluorescence intensity was measured using a microplate reader. After the experiment, the intestinal segments were removed, and the surface area of each segment was recorded. The Papp values were calculated.

Balb/c mice (8 weeks old) were fasted overnight. Subsequently, the mice were anesthetized, and a midline laparotomy was performed to expose the proximal small intestine. Approximately 2 cm intestinal segments were carefully isolated. C6-labeled vesicles (C6 concentration: 1.0 μg/mL) were injected into these loops, which were then ligated to form closed systems. After a 3 h incubation period, the intestinal segments were excised and subjected to sequential processing: initial rinsing with K-R solution, fixation in 4% paraformaldehyde, and dehydration through 30% sucrose immersion. Tissue specimens were embedded in optimum cutting temperature medium and sectioned at 10 μm thickness using a Leica cryostat (Germany). Nuclear counterstaining was performed with DAPI prior to fluorescence imaging analysis using an Olympus confocal laser scanning microscope (CLSM, Japan).

Pharmacokinetics of Cef-loaded vesicles in vivo

Balb/c mice (8 weeks old) were fasted overnight and orally administered free Cef, Lip@Cef, mEV@Cef or mEV-Lip@Cef (Cef: 35 mg/kg). Serial blood samples (0.2 mL) were obtained via retro-orbital puncture at specified intervals and immediately transferred to heparinized tubes. The quantification of Cef content in the plasma was performed utilizing HPLC. The chromatographic separation employed a C18 column (250 mm × 4.6 mm, 5 μm) maintained at 40°C. An isocratic mobile phase containing water-methanol-formic acid (100:10:0.1, v/v/v) and methanol (82:18, v/v) was delivered at 1.0 mL/min. Drug detection was achieved through ultraviolet absorbance monitoring at 286 nm.

In vivo pharmacodynamics studies in lung infection model with Klebsiella pneumoniae

A murine pneumonia model was established in Balb/c mice (8 weeks old) by intranasal inoculation with 1 × 10⁷ CFU of Klebsiella pneumoniae suspended in 50 μL of medium. Therapeutic interventions were initiated 24 h post-infection, with three oral administrations (24, 36, and 48 h) of either free Cef (20 mg/kg) or Cef-loaded vesicles at equivalent dosage. Lung tissue processing and analysis were conducted 72 h post-infection. Collected pulmonary tissues were homogenized in sterile PBS and subjected to serial dilutions. Quantitative bacteriological analysis was performed by culturing homogenate on selective MacConkey agar, followed by 12 h incubation at 37 °C for CFU enumeration. Proinflammatory cytokine concentrations (IL-6 and TNF-α) were determined in lung homogenates using Elisa kits. For morphological analysis, tissue specimens were fixed in 4% paraformaldehyde, paraffin-embedded, and sectioned for histological staining. Sections underwent hematoxylin and eosin (H&E) and Gram staining, with microscopic examination conducted using light-field optics.

In vivo pharmacodynamics studies in bacteremia model with Klebsiella pneumoniae

Bacteremia was induced in Balb/c mice (8 weeks old) by intravenous injection with 1 × 10⁷ CFU Klebsiella pneumoniae. Therapeutic interventions were initiated at two time points (2 and 8 h) post-infection, with experimental groups receiving either free Cef (20 mg/kg) or equivalent Cef-loaded vesicles via oral gavage. To quantify bacterial dissemination, peripheral blood specimens were obtained at 24 h post-infection. Each 0.1 mL blood aliquot underwent initial tenfold dilution with sterile PBS, followed by serial decimal dilutions for plating on selective MacConkey agar. Concurrently, hepatic and splenic tissues were aseptically excised, weighed and homogenized in 1 mL PBS. Tissue homogenates were processed through successive tenfold dilutions before culturing on MacConkey agar. All culture plates underwent overnight incubation at 37 °C. Colonies were counted to calculate the CFU. The liver and spleen were harvested for H&E staining histological examination.

High-fat diet (HFD) mouse model

Balb/c mice (3 weeks old) were divided into three groups receiving PBS, free Cef or mEV-Lip@Cef via oral gavage for 5 d. Subsequently, all animals were transitioned to HFD (XTHF60-1, Jiangsu Collaborative Biotechnology Engineering Co., Ltd., Nanjing, China) from week 7 to week 10 of age. Longitudinal body weight measurements were recorded at weekly intervals during the HFD phase. Metabolic characterization was performed at week17 through an overnight-fasted intraperitoneal glucose tolerance test (IPGTT), where animals received 1 g/kg glucose solution via intraperitoneally injection. Blood glucose concentrations were quantified using a glucometer at predetermined time points after glucose injection. Terminal procedures were conducted at week 19, with subsequent organ collection for biochemical and histopathological evaluation. Excised hepatic tissues were immediately weighed and processed into homogeneous suspensions for quantitative analysis of total cholesterol content using a commercial cholesterol assay kit following manufacturer specifications. The livers were harvested for histological examination via H&E staining.

Fecal microbiota transplantation (FMT)

Donor Balb/c mice received daily oral gavage of PBS, free Cef, or mEV-Lip@Cef formulations for 5 d. Fecal specimens were collected from the donor mice under sterile conditions for FMT. Individual fecal samples underwent sterile processing involving homogenization in PBS followed by gravitational sedimentation. The supernatants were immediately aliquoted under controlled conditions to recipient mice. Following FMT administration, recipient mice were subjected to HFD for 13 weeks and analyzed as described above.

Intestinal infection model with Escherichia coli

Balb/c mice (8 weeks old) were pretreated with free Cef or mEV-Lip@Cef via oral administration daily for 5 d, while control animals were administered equivalent volumes of PBS via the same route. Subsequently, all experimental groups were subjected to intraperitoneal challenge with Escherichia coli (1 × 10⁸ CFU per injection) on days 5 and 9, with analysis on day 13. Fecal, caecum, and colon specimens were aseptically collected, weighed and immediately homogenized in sterile PBS. The resulting homogenates underwent serial dilutions in PBS followed by plating on selective MacConkey agar medium for bacterial quantification. Cultured plates were maintained at 37 °C overnight before enumeration of CFU, with final counts normalized to tissue mass. The colons were harvested for histological examination via H&E staining.

Data analysis

Experimental data are presented as mean ± standard deviation values (Mean ± SD). Statistical analyses were performed using IBM SPSS Statistics 23.0. Comparisons between two experimental groups were conducted using the two-tailed Student’s t test. For multiple comparisons, one-way analysis of variance (ANOVA) was employed, followed by post hoc analysis with Tukey’s honest significant difference (HSD) test. P < 0.05 were considered statistically significant.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Y.Y. conceived the study, designed the experimental strategy, and wrote the manuscript. Y.X., Z.Y., X.L., J.G., M.X., and J.C. conducted the experiments and acquired the data. R.T.W. reviewed and revised the manuscript. L.L. contributed the experimental design and reviewed and edited the manuscript.

Peer review

Peer review information

Nature Communications thanks Kui Zhu,Driton Vllasaliu, and Jingyuan Wen for their contribution to the peer review of this work. A peer review file is available.

Data availability

All data are available within the article, supplementary information, and the associated “Source Data” file. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-68082-9.