Engineering oral celastrol-polysaccharide supramolecular nanoassemblies across intestinal barriers for the intervention of nonalcoholic steatohepatitis

Abstract

The development of effective oral therapies for nonalcoholic steatohepatitis (NASH) remains a critical unmet need in clinical practice. Celastrol (CEL), a potent natural compound, is a promising candidate for NASH due to its lipid-modulating, anti-inflammatory properties, and antioxidative properties. However, its clinical translation is severely hindered by poor oral bioavailability and a narrow therapeutic window stemming from significant toxicity. To overcome these limitations, we present the design and fabrication of novel, carrier-free CEL nanoparticles stabilized through natural polysaccharide self-assembly. This innovative formulation strategy is aimed at substantially enhancing the oral bioavailability of CEL, thereby augmenting its therapeutic efficacy in NASH while mitigating associated adverse effects. In this study, we developed a polysaccharide-CEL supramolecular depot comprising 42 distinct nanoassemblies for oral delivery in NASH, identifying chondroitin sulfate (CS) as the optimal polysaccharide adjuvant. The resulting CS/CNA formulation markedly improved oral absorption, with a 3.12-fold increase in observed systemic exposure based on AUC_(0-t) compared with free CEL. Inhibitor-based mechanistic studies suggested that CS/CNA transport across intestinal epithelial cells may involve multiple endocytosis pathways and partial lysosomal escape. In a murine model of NASH, CS/CNA demonstrated potent therapeutic efficacy by modulating hepatic inflammation and reducing lipid accumulation. Critically, the nanoassembly exhibited a favorable safety profile, mitigating the toxicity associated with free CEL. By enhancing therapeutic efficacy while reducing adverse effects, this strategy effectively widens the therapeutic window for celastrol. This work establishes CS/CNA as a promising oral therapeutic candidate and provides a robust platform for advancing the clinical potential of CEL for NASH management.

Graphical abstract

1. Introduction

Nonalcoholic steatohepatitis (NASH) represents a crucial stage of nonalcoholic fatty liver disease (NAFLD), affecting approximately 5% of individuals worldwide [[1], [2], [3], [4], [5]]. Without timely intervention, it may progress to advanced liver diseases such as cirrhosis and liver cancer [4]. Numerous innovative drugs, such as Resmetirom and Semaglutide, have been developed to combat NASH, but challenges remain [[6], [7], [8], [9], [10]]. On the one hand, the pathogenesis of NASH is complex, meaning that the existing drug may not adequately address the pathological characteristics of different patient types [11,12]. On the other hand, some drugs are primarily administered through cumbersome injection [6]. Given NASH's chronic nature, oral medications offer better patient compliance and convenience compared to other routes of administration, facilitating long-term treatment for NASH patients [6]. Therefore, the development of new oral therapeutic options will provide NASH patients with more diversified and convenient treatment choices.

Celastrol (CEL), a natural compound extracted from the root bark of Tripterygium wilfordii, has shown significant medicinal potential in regulating lipid metabolism, reducing fat accumulation, promoting weight loss, and protecting the liver, thereby improving NASH [[13], [14], [15], [16], [17], [18], [19]]. However, the clinical application of CEL is limited by its poor water solubility, low oral bioavailability, and potential toxicity [13,[20], [21], [22]]. Recent CEL-based metal nanomedicines, ATP-activatable chelates, and copper–CEL nanohybrids further highlight the growing interest in developing safer and more efficient CEL therapeutic systems [[23], [24], [25]]. Therefore, an effective oral delivery system is needed to overcome these limitations and facilitate the clinical translation of CEL.

In recent years, natural polysaccharides have shown great potential in the field of nanomedicine due to their excellent biocompatibility and low toxicity [[26], [27], [28], [29]]. In this study, polysaccharides were selected as the guest molecular chaperones for CEL based on the following four reasons: Firstly, polysaccharides contain numerous hydrophilic groups such as hydroxyl, carboxyl, and amino groups, which not only provide good water solubility but also facilitate supramolecular self-assembly with guest molecules through hydrogen bonding, forming stable nanostructures [[30], [31], [32]]. This promotes the ordered arrangement of molecules, favoring the dispersion and dissolution of CEL in vivo. Secondly, polysaccharides may interact with intestinal mucus and thereby prolong mucosal retention after oral administration [33,34]. Although chitosan is the best-known cationic mucoadhesive polysaccharide, anionic polysaccharides such as chondroitin sulfate may also interact with mucus through non-electrostatic mechanisms, including hydrogen bonding between its hydroxyl, carboxylate, sulfate, and acetamido groups and mucin glycoproteins, as well as possible chain entanglement within the mucus gel matrix [35,36]. Thirdly, polysaccharides can serve as surfactants, adjusting the size, shape, and charge of nanoparticles depending on dosage, leading to uniformly sized and regularly shaped nanoparticles [37]. This may further influence cellular uptake and in vivo biodistribution. Lastly, the assembly of polysaccharides with CEL can reduce the direct contact between CEL and the organism, thereby lowering its potential toxicity [29].

To identify the optimal polysaccharide companion for CEL self-assembly, we selected seven representative natural polysaccharides, including chitosan, astragalus polysaccharide, dextran, chondroitin sulfate, fucoidan, ginseng polysaccharide, and sodium alginate, to construct a structurally diverse screening library. These candidates were chosen to cover a broad physicochemical spectrum in terms of surface charge, functional groups, and molecular weight. Specifically, chitosan was included as a positively charged polysaccharide containing amino groups; chondroitin sulfate, fucoidan, and sodium alginate were selected as negatively charged polysaccharides bearing sulfate and/or carboxylate groups; and dextran, astragalus polysaccharide, and ginseng polysaccharide were included as relatively neutral polysaccharides rich in hydroxyl groups. These structural differences were expected to influence CEL–polysaccharide self-assembly, nanoparticle stability, surface charge, and potential interactions with the intestinal mucus barrier. By formulating these seven candidates with CEL at different mass ratios, we generated a supramolecular screening library comprising 42 distinct nanoassemblies, thereby increasing the likelihood of identifying an optimal polysaccharide companion for oral CEL delivery.

Based on this rationally designed screening strategy, we identified chondroitin sulfate (CS) as the most suitable companion polysaccharide for CEL and further optimized its assembly ratio to obtain CS–CEL nanoassemblies (CS/CNA) with favorable physicochemical properties for oral administration (Fig. 1A). We then systematically investigated the structural characteristics, gastrointestinal stability, intestinal retention, oral bioavailability, and absorption mechanisms of CS/CNA. Finally, the therapeutic efficacy and safety of CS/CNA were evaluated in a murine model of NASH, with particular attention to liver morphology, pathological alterations, lipid accumulation, and inflammatory markers (Fig. 1B). This work provides a polysaccharide-guided supramolecular nanoassembly strategy to improve the oral delivery and therapeutic potential of CEL for NASH intervention.

Fig. 1.

(A) Schematic outline of the self-assembly of polysaccharide aqueous solution with CEL; (B) Schematic outline of oral delivery and application in hepatic NASH therapy.

2. Materials and methods

2.1. Materials and reagents

Celastrol (CEL, purity ≥98%), chitosan (molecular weight: 55000 Da, degree of deacetylation: ≥95%, purity ≥98%), astragalus polysaccharide (molecular weight: 5650 Da, purity ≥70%), dextran (molecular weight: 1500 Da, purity ≥95%) and sodium alginate (molecular weight: 30000 Da, M/G = 1:2, purity ≥90%) were purchased from Macklin Chemicals (Shanghai, China). Fucoidan (molecular weight: 18000 Da, purity ≥95%) was purchased from Aladdin Chemicals (Beijing, China).

Ginseng polysaccharide (molecular weight: 63000 Da, purity ≥98%) was bought from Xi An Tongze Biotech (Xian, China). Chondroitin sulfate (molecular weight: 75000 Da, <0.24 sulfates per disaccharide, purity ≥95%) was purchased from Sangon Biotech (Shanghai, China). Cy5-amine and resmetirom (purity ≥99.90%) was bought from MedChemExpress (Shanghai, China). Emodin (purity ≥98.0%) was bought from Beyotime biotechnology (Shanghai, China). The following primary antibodies were used: Anti-β-Actin (HRP-66009, Proteintech Group, United States), anti-TNF-α (SC-1351, Santa Cruz Biotech, United States), anti-ACTA2 (R23450, ZEN-BIOSCIENCE, China), anti-IL-6 (PB0060, Boster Biological Technology, United States). HFMCD (60 kcal% fat, Methionine 0.1% and choline deficiency) mouse feed was purchased from Jiang Su Xietong Pharmaceutical Bio-engineering Co. (Nanjing, China). All other chemicals were of analytical grade and used as received. Deionized water (18.2 MΩ, 25°C) from a Milli-Q Plus water purification system (Millipore) was used throughout the study.

2.2. Animals and cells

Healthy male C57BL/6 mice (8-10 weeks old, weighing 15-20 g) were obtained from the Animal Centre of Chongqing Medical University, while healthy male Sprague-Dawley rats (6-8 weeks old, weighing 200 g) were purchased from Hunan S&T Biotech Co., Ltd (China). All mice were housed in controlled experimental animal facilities, under a 12-h light/dark cycle, maintained at a temperature of 24 ± 2°C and humidity levels of 50 ± 10%. They were provided with sufficient food and water throughout the experiment.

All animal experiments were conducted in strict accordance with the guidelines provided by the National Research Council's Guide for the Care and Use of Laboratory Animals. All procedures involving animals were approved by the Institutional Animal Care and Use of Chongqing Medical University (IACUC-CQMU). Approval number: IACUC-CQMU-2023-0375.

Caco-2 and HepG2 cells were purchased from Wuhan Pricella Biotechnology Co., Ltd. (China). The cells were cultured in DMEM supplemented with 10% FBS, 100 U/mL penicillin, and 10 mg/mL streptomycin, and maintained at 37°C in a humidified atmosphere of 5% CO₂.

2.3. Preparation of polysaccharide supramolecular nanoassemblies

Celastrol was dissolved in DMSO to form a 10 mg/mL stock solution, and polysaccharide solutions of varying concentrations (0.2, 1, 2.5, 5, 10, 20 mg/mL) were prepared separately. A 20 μL aliquot of Celastrol stock solution was then added dropwise to 2 mL of polysaccharide solution and stirred at room temperature for 30 min. The reaction mixture was dialyzed in water to remove unbound Celastrol, polysaccharide, and excess DMSO, yielding the self-assembled polysaccharide nanoparticles. The resulting nanoparticle suspension was freeze-dried and stored at 4°C for subsequent experiments.

2.4. Characterization of polysaccharide supramolecular nanoassemblies

The zeta potential, dispersion coefficient, and particle size of the nanoparticles were measured using dynamic light scattering (DLS) at a temperature of 25°C. Transmission electron microscopy (TEM) images were obtained by placing nanoparticle suspensions on a copper grid, removing excess solution, staining with 10 μM phosphotungstic acid, and air-drying at room temperature.

For the characterization of CS/CNA, X-ray diffraction (XRD) spectra were collected for free Celastrol, CS, CS/CNA, and physical mixtures of CEL/CS. These spectra were obtained using 40 kV and 40 mA, with scans from 3° to 80°, using copper as the target material.

The drug loading efficiency (DLE%) and encapsulation efficiency (EE%) of the CS/CNA were determined by measuring the CEL content before dialysis (W_a) using a UV spectrophotometer and weighing the total mass of CS/CNA (W_t) after freeze-drying. The sample was redissolved in 2 mL of ultrapure water, and the CEL content after redissolution (W_b) was again measured by UV spectrophotometer. The DLE% and EE% were calculated using the respective formulas:

$$DLE\%=\frac{W_b}{W_t}\times 100\%$$

$$EE\%=\frac{W_b}{W_a}\times 100\%$$

To evaluate the molecular interactions involved in CS/CNA self-assembly, CS/CNA was incubated with PBS, 10% SDS, or 8 M urea at 37°C for 2 h, followed by DLS analysis. FTIR spectra of free CEL, CS, CEL/CS physical mixture, and CS/CNA were recorded over 4000–500 cm⁻¹.

2.5. Physicochemical stability evaluation of CS/CNA

To evaluate the stability of CS/CNA in gastrointestinal fluid, artificial intestinal fluid (SIF), artificial gastric fluid (SGF), and PBS were prepared and mixed with CS/CNA at a 1:1 (v/v) ratio. The mixture was incubated at 37°C for 2 h in SGF, and for 4 h in both SIF and PBS. Changes in particle size and polydispersity index (PDI) were measured using DLS. Morphological changes and particle distribution were observed through TEM after incubation.

For room temperature stability, CS/CNA were stored at 25°C for one week. Daily measurements of particle size and PDI were conducted using DLS to monitor stability over time.

To further evaluate physiological stability, CS/CNA was incubated in buffers with different pH values, including pH 1.2, pH 6.8, and pH 7.2, at 37°C for 6 h. At predetermined time points, hydrodynamic diameter and PDI were measured by DLS. To assess ionic strength stability, CS/CNA was incubated in NaCl solutions at 150 mM and 500 mM, and particle size and PDI were measured by DLS.

2.6. Molecular dynamics simulations

Molecular dynamics simulations were performed using AMBER 22 software [38]. The initial system contained representative CS tetrasaccharide fragments and CEL molecules at a CS tetrasaccharide fragment: CEL molar ratio of 2:1. The full-length CS polymer was simplified as discrete tetrasaccharide fragments composed of alternating β-D-glucuronic acid and N-acetyl-D-galactosamine disaccharide units. To represent the sulfated CS motif under model simplification, one sulfate group was introduced at the C4-hydroxyl position of the galactosamine residue. The CS glycan structures were parameterized using the GLYCAM06 force field, whereas CEL topology and partial charges were assigned using GAFF2 and the AM1-BCC charge model, respectively. Deprotonated carboxylate and sulfate groups were neutralized by adding Na⁺ counterions, and the system was solvated in a TIP3P water box.

After energy minimization, the system was heated from 0 to 298.15 K over 200 ps, equilibrated under NVT and NPT conditions, and subjected to a 150 ns NPT production simulation. Long-range electrostatic interactions were treated using the Particle Mesh Ewald method with a 10 Å non-bonded cutoff. Hydrogen bonds were constrained using the SHAKE algorithm, and the temperature was maintained using a Langevin thermostat. The integration timestep was 2 fs, and trajectories were saved every 10 ps for post-analysis [39].

2.7. In vitro release assay of CS/CNA

To assess the cumulative release of CS/CNA in vitro, samples were gently stirred in SIF and SGF at 37°C using a magnetic stirrer. Specifically, 2 mL of CS/CNA were placed in a dialysis bag, which was then immersed in 20 mL of the SIF or SGF at 37°C and stirred at 150 rpm. Samples were collected at 0.5, 1, 1.5, and 2 h during incubation in SGF. For the SIF, samples were collected at 0.5, 1, 1.5, 2, 2.5, 3, 3.5, and 4 h. The content of CEL in each sample was measured using a UV spectrophotometer.

2.8. In vivo distribution and retention studies

Cy5-labeled CS/CNA were prepared for in vivo distribution and retention studies. A solution of 2 mL of 10 mg/mL CS was activated with 10 mg of NHS and EDC and stirred at 300 rpm for 30 min. After activation, 500 μL of 0.1 mg/mL Cy5-amine in DMSO was added. The reaction was protected from light for 24 h, followed by dialysis for 24 h to obtain Cy5-CS. Cy5-CS/CNA were then prepared using the previously described self-assembly method.

Thirty-six male C57BL/6 mice were randomly assigned to two groups: free Cy5 and Cy5-CS/CNA. Mice were administered free Cy5 or Cy5-CS/CNA orally (40 μg/kg, calculated based on Cy5 content). Following administration, mice were scanned using a live imaging system (Vieworks) at 0.5, 1, 3, 6, 12, and 24 h. Additionally, mice were sacrificed at these time points to collect images of their gastrointestinal tract and organs (heart, liver, spleen, lungs, and kidneys) using the same imaging system. Fluorescence imaging was performed with an excitation wavelength of 649 nm and an emission wavelength of 670 nm.

2.9. Pharmacokinetic analysis

The male Sprague-Dawley (SD) rats were randomly divided into two groups: CEL and CS/CNA. CEL and CS/CNA were administered orally (10 mg/kg CEL). Blood samples (200-300 μL) were collected from the orbital vein at 0.25, 0.5, 1, 2, 4, 6, 8, 12, and 24 h post-administration. Blood samples were centrifuged at 4000 rpm for 10 min at 4°C, and plasma was stored at −80°C for further analysis.

To prepare the standard curve, CEL was dissolved in methanol to create a 10 mg/mL solution, which was mixed with blank plasma to obtain concentrations ranging from 0.0781 to 10 μg/mL (0.0781, 0.1563, 0.3125, 0.625, 1.25, 2.5, 5, and 10 μg/mL). Emodin (20 μg/mL in methanol) was used as an internal standard. 10 μL of internal standard solution was mixed with 100 μL of rat plasma and 400 μL of ethyl acetate and vortexed for 3 min. Subsequently, the mixture was desolventized in a vacuum centrifugal concentrator, 90 μL of ethanol was added and vortexed for 3 min to reconstitute and centrifuged in a high-speed refrigerated centrifuge at 13500 rpm at 4°C for 10 min. Finally, 70 μL of supernatant was transferred to an inner insert tube for sample analysis. Plasma samples were analyzed by liquid chromatography (HPLC, Shimadzu). Representative chromatograms of the samples are shown in the Supplementary S4.

Pharmacokinetic parameters were calculated using non-compartmental analysis. AUC_(0-t) was calculated using the linear trapezoidal method based on the observed plasma concentration–time profiles. The terminal elimination rate constant (λz) was estimated from the terminal log-linear phase using at least three quantifiable terminal time points when applicable. AUC_(0-∞) was calculated as AUC_(0-t) + C_last/λz. Because several terminal-phase extrapolated parameters showed high variability, particularly in the CS/CNA group, the pharmacokinetic interpretation mainly focused on observed parameters, including AUC_(0-t), Cmax, and Tmax.

2.10. Cellular uptake

Caco-2 cells were seeded in a 24-well plate at a density of 1 × 10⁵ cells per well, and 1 mL of complete medium was added. The cells were cultured in a 37°C, 5% CO2 incubator for 48 h. Once cell confluency reached 80-90%, free Cy5 and Cy5-CS/CNA were added and incubated for 1, 3, 6, 12, and 24 h at 37°C. After incubation, cells were harvested, and the uptake of CS/CNA was analyzed using flow cytometry.

Additionally, cells could be fixed with 4% paraformaldehyde and stained with DAPI for 15 min. The uptake of CS/CNA was analyzed via confocal microscopy, and fluorescence intensity was quantified using ImageJ software.

2.11. Transport mechanism assay of CS/CNA in Caco-2 cells

A Caco-2 monolayer model was constructed for transport studies. Caco-2 cells were seeded at a density of 2 × 10⁵ in 24-well plates with Transwell inserts and cultured for 21 days. The trans-epithelial electrical resistance (TEER) of the monolayer was measured every three days, and when TEER values exceeded 500 Ω cm², the model was used for transmembrane transport experiments.

To test the intestinal permeability of CS/CNA, 500 μL of culture medium containing free Cy5 or Cy5-CS/CNA was added to the upper chamber of the Transwell, and the system was incubated for 12 h. Fluorescence intensity at 648 nm in the lower chamber was measured using a microplate reader, and permeability was calculated using the appropriate formula.

To explore the endocytosis and exocytosis pathways, logarithmically growing Caco-2 cells were treated with pharmacological inhibitors commonly used to probe endocytic or intracellular trafficking pathways for 1 h before incubation with Cy5-CS/CNA for 6 h. After incubation, cells were analyzed by flow cytometry. Additionally, cells were incubated with fluorescent probes targeting lysosomes, the Golgi apparatus, and the endoplasmic reticulum, and Hoechst staining was used for nuclei. The co-localization of CS/CNA with these organelles was observed using a laser confocal microscope.

2.12. Biocompatibility evaluation

To evaluate the short-term safety of CS/CNA, male C57BL/6 mice were randomly divided into five groups: control, low-dose CEL (4 mg/kg), low-dose CS/CNA (4 mg/kg CEL equivalent), high-dose CEL (20 mg/kg), and high-dose CS/CNA (20 mg/kg CEL equivalent). After seven days of acclimation, mice were orally administered the corresponding formulations once daily for 7 consecutive days. Body weight and general condition were monitored during the experiment. At the end of the treatment, mice were sacrificed, and blood and major organs were collected. Hematological and serum biochemical parameters were analyzed, and major organs were subjected to hematoxylin and eosin (H&E) staining for histopathological examination.

To further assess the repeated-dose safety of CS/CNA, healthy male C57BL/6 mice were randomly divided into four groups: vehicle, low-dose CS/CNA (4 mg/kg CEL equivalent, L-CS/CNA), medium-dose CS/CNA (10 mg/kg CEL equivalent, M-CS/CNA), and high-dose CS/CNA (20 mg/kg CEL equivalent, H-CS/CNA). Mice were orally administered vehicle or CS/CNA at different doses for 30 consecutive days under a normal diet. Body weight was recorded weekly. At the end of the experiment, blood samples and major organs, including the heart, liver, spleen, lung, kidney, and brain, were collected. Gross organ morphology was photographed, and organ coefficients were calculated as organ weight/body weight × 100%. Serum biochemical parameters and hematological indicators were measured. Major organs were fixed, sectioned, and stained with H&E for histopathological evaluation.

2.13. Pharmacodynamics study

Male C57BL/6 mice were randomly assigned to five groups: control, HFMCD, HFMCD-CEL, HFMCD-CS/CNA, and HFMCD-resmetirom (with respective treatments). The control group was fed a normal diet, while the experimental groups received a high-fat methionine- and choline-deficient (HFMCD) diet to induce non-alcoholic steatohepatitis (NASH). After four weeks, animals in the treatment groups were orally administered solvent, CEL, CS/CNA (10 mg/kg CEL equivalent), or resmetirom (10 mg/kg). Weight changes were recorded daily.

At the eighth week, glucose and insulin tolerance tests were performed, and blood glucose levels were monitored at 0, 15, 30, 60, 90, and 120 min. Mice were then sacrificed, and serum and organs were collected for biochemical analysis of ALT, AST, TG, and TC. Liver sections were stained using HE, Masson, Sirius red, and Oil Red O to evaluate liver damage. Western blot analysis was performed to assess cytokine (TNFα, IL-6) and ACTA2 expression.

2.14. Data and statistical analysis

Data are presented as mean ± SEM. All experiments were repeated at least three times to ensure reproducibility. Data were statistically analyzed using Prism 8.0 (GraphPad Software, USA) using the Two-tailed unpaired t-test or one-way ANOVA. Differences were considered statistically significant if p < 0.05.

3. Results

3.1. Screening and physicochemical characterization of polysaccharide-CEL nano-assemblies

To identify a suitable polysaccharide companion for CEL self-assembly, we constructed a screening library consisting of 42 polysaccharide-CEL nanoassemblies using seven representative natural polysaccharides. The resulting nanoassemblies were systematically characterized in terms of particle size, PDI, zeta potential, and TEM morphology (Fig. 2 and Supplementary Table S1).

Fig. 2.

Screening and physicochemical characterization of polysaccharide-CEL nanoassemblies. (A) Hydrodynamic diameter and polydispersity index (PDI) of polysaccharide-CEL nanoassemblies prepared from seven representative polysaccharides at different CEL: polysaccharide mass ratios. The left y-axis indicates particle size, and the right y-axis indicates PDI. (B) Zeta potential of the corresponding polysaccharide-CEL nanoassemblies. (C) Representative TEM and magnified TEM images showing the morphology of nanoassemblies formed with different polysaccharides. Scale bars: 500 nm in TEM images and 100 nm in magnified TEM images. Data are presented as mean ± SEM, n = 3. Detailed particle size, PDI, zeta potential, and morphology of all 42 nanoassemblies are summarized in Supplementary Table S1.

The particle size and surface charge of the nanoassemblies varied substantially depending on the polysaccharide type and the CEL:polysaccharide mass ratio. Ginseng polysaccharide- and chondroitin sulfate-based nanoassemblies generally formed well-defined nanosized particles, whereas fucoidan-, chitosan-, and sodium alginate-based formulations tended to form larger particles or pronounced aggregates. Astragalus polysaccharide-based nanoassemblies showed relatively large size variation among different mass ratios. Zeta potential analysis showed that most nanoassemblies were negatively charged, while chitosan-based formulations exhibited positive surface charges, ascribed to the protonated amino groups of chitosan (Fig. 2B). TEM observation further revealed distinct morphological differences among the nanoassemblies. Ginseng polysaccharide- and astragalus polysaccharide-based assemblies formed particles with visible size heterogeneity, whereas dextran-, chitosan-, and sodium alginate-based formulations showed prominent structural aggregation or irregular, amorphous morphologies. In contrast, chondroitin sulfate-based nanoassemblies (CS/CNA) formed well-dispersed and highly uniform spherical nanoparticles (Fig. 2C, Supplementary S1).

To further expand the screening dimensions beyond physicochemical characterization, hemolysis assays were performed as an in vitro hemocompatibility evaluation for free CEL and representative polysaccharide-CEL nanoassemblies at CEL-equivalent concentrations (Supplementary S2). CS-CEL nanoassemblies (CS/CNA) showed low hemolytic activity across the tested concentration range, whereas some candidate formulations exhibited increased hemolysis at higher concentrations. These results suggested that CS/CNA possessed favorable hemocompatibility, providing additional biological support for the selection of CS as the polysaccharide companion for CEL self-assembly.

Overall, CS/CNA was selected as the optimal formulation based on an integrated evaluation of physicochemical properties, TEM morphology, and in vitro hemocompatibility. Although several ginseng polysaccharide- and astragalus polysaccharide-based nanoassemblies showed comparable nanoscale sizes, they displayed broader size distributions, less uniform morphology, or partial aggregation. In contrast, CS/CNA exhibited a small and relatively uniform particle size, low PDI, a moderately negative zeta potential (−10 to −20 mV), well-dispersed spherical morphology, and favorable hemocompatibility. These combined features suggested that CS/CNA had a more favorable profile for oral delivery than the other polysaccharide-CEL nanoassemblies [[31], [32], [33]]. Detailed physicochemical properties and morphology of all 42 nanoassemblies are summarized in Supplementary Table S1.

3.2. Structure and Oral Absorption Potential of CS/CNA

The X-ray diffraction (XRD) analysis (Fig. 3A) revealed a strong and sharp diffraction peak for CEL at 10-20°, indicating its crystal structure, with prominent peaks at 9.2367°, 14.51832°, and 16.74011°. In contrast, the CS/CNA did not display any distinct peaks, suggesting that the assembled CS/CNA were amorphous. The disappearance of the characteristic crystalline diffraction peaks of CEL indicated that CEL was transformed into an amorphous or molecularly dispersed state within CS/CNA. A high encapsulation efficiency is crucial for drug safety and stability, while a high drug loading efficiency is directly correlated with the ability to achieve effective clinical dosages within reasonable administration volumes and frequencies. Therefore, we evaluated both encapsulation and drug loading efficiencies of CS/CNA at different feeding ratios. A feeding ratio of 1:100 yielded the maximum drug loading efficiency (DLE) of 17.463 ± 0.120% and encapsulation efficiency (EE) of 83.914 ± 2.052% (Fig. 3B and C). Additionally, DLS results demonstrated that both the particle size and zeta potential of CS/CNA (1:100) remained stable over a 7-day period, confirming their good in vitro storage stability (Fig. 3D). MD simulations were employed to visualize the interactions between CS and CEL, as well as to assess the stability of the self-assembled nanostructures. The results indicated that the CS/CNA system underwent a spontaneous aggregation process (Fig. 3E, F, G, H, Supplementary S3A, B). Initially, CEL and CS were randomly distributed within the continuous phase. However, after 150 ns, they tightly connected to form spherical aggregates. Notably, the CEL molecules were firmly embedded within the CS molecules through hydrogen bonding and hydrophobic interactions. Interaction-disruption assays and FTIR analysis were further performed to support the MD-predicted CS–CEL interactions. Compared with CS/CNA in PBS, incubation with 10% SDS or 8 M urea caused broadened or multi-peak particle size distributions, indicating partial disruption of the nanoassembly structure (Supplementary S4A, B, C). FTIR analysis showed changes in O–H/N–H stretching and C=O vibration bands after CS/CNA formation, further supporting the involvement of hydrophobic interactions and hydrogen bonding in CS/CNA self-assembly (Supplementary S4D).

Fig. 3.

Structure and Oral Absorption Potential of CS/CNA. (A) XRD patterns of free CEL, chondroitin sulfate, CEL/CS physical mixture, and CS/CNA; (B) Encapsulation efficiency and (C) drug loading efficiency of CS/CNA prepared at different CEL:CS mass ratios (n = 9); (D) Changes in hydrodynamic diameter and PDI of CS/CNA during storage at 25°C for 7 days (n = 3); (E) Representative conformation of the CS/CNA system after 150 ns molecular dynamics simulation; (F) Root mean square deviation (RMSD), (G) number of hydrogen bonds, and (H) solvent-accessible surface area (SASA) of the CS/CNA system during the 150 ns simulation; (I) In vitro CEL release profiles of CS/CNA prepared at different CEL:CS mass ratios in simulated gastric fluid (SGF) and (J) the corresponding area under the release curve (AUC) (n = 3); (K) In vitro CEL release profiles of CS/CNA prepared at different CEL:CS mass ratios in simulated intestinal fluid (SIF) and (L) the corresponding AUC (n = 3); (M) Schematic illustration of the FRET-based assay used to evaluate the structural integrity of CS/CNA; (N) Quantitative analysis of FRET signals after incubation of CS/CNA in PBS, SGF, and SIF (n = 3); (O) Hydrodynamic diameter distribution of CS/CNA after incubation in PBS, SGF, and SIF; (P) Representative TEM and magnified TEM images of CS/CNA after incubation in PBS, SGF, and SIF. Scale bars: 500 nm in TEM images and 200 nm in magnified TEM images. Data are presented as mean ± SEM. Statistical significance was analyzed by one-way ANOVA. *p < 0.05, ***p < 0.001.

To further assess the oral delivery potential of CS/CNA, we evaluated the stability and drug release of CS/CNA in simulated gastric fluid (SGF) and simulated intestinal fluid (SIF). CS/CNA at a 1:100 ratio showed the lowest drug release rates in both SGF and SIF, indicating that their structure was more resistant to degradation in the gastrointestinal environment (Fig. 3I, J, K, L). This structural stability enhances their potential for absorption in the small intestine. The interaction of DiI and DiO effectively produces the FRET effect due to the highly overlapping spectra of emission and absorption wavelengths [40]. The stability of the materials can be better assessed through the FRET effect (Fig. 3M). The results indicated that the CS/CNA encapsulated with DiI and DiO exhibited a significant FRET effect under PBS, while the FRET effect remained stable during incubation with SGF and SIF (Fig. 3N, Supplementary S5). DLS analysis showed a slight increase in particle size for CS/CNA at the 1:100 ratio after incubation in SGF and SIF, while TEM confirmed that there were no significant changes in nanoparticle morphology (Fig. 3O and P). This suggested that gastrointestinal fluids did not disrupt the nanostructure of the CS/CNA. Furthermore, CS/CNA maintained nanoscale size under pH 1.2, 6.8, and 7.2 conditions, and showed no obvious changes in particle size or PDI in 150 mM and 500 mM NaCl solutions, supporting its colloidal stability under physiologically relevant pH and ionic-strength conditions (Supplementary S6).

These findings collectively indicated that the self-assembled CS/CNA were stable, exhibited minimal drug release in gastrointestinal fluids, and may be favorable for oral delivery and subsequent intestinal absorption.

3.3. CS/CNA improves oral absorption and preferential hepatic accumulation after oral administration

To evaluate the in vivo biodistribution and intestinal retention of orally administered CS/CNA in mice, Cy5-CS was self-assembled with CEL to obtain Cy5-CS/CNA. Male C57 mice were orally administered the Cy5-CS/CNA solution along with free Cy5 (Fig. 4C). In vivo imaging results showed that, after 3, 6, and 12 h of administration, the average fluorescence intensity in the free Cy5 group was significantly lower than in the Cy5-CS/CNA group (Fig. 4A). Additionally, the retention of Cy5-CS/CNA in the isolated intestinal tract of the mice was notably longer than that of the free Cy5 group, with increased accumulation in the jejunum (Fig. 4B–D, E, F, G). Furthermore, the fluorescence intensity in the liver was higher in the Cy5-CS/CNA group than in the free Cy5 group, indicating preferential hepatic accumulation after oral administration (Fig. 4H and I). However, these fluorescence imaging data do not distinguish passive first-pass hepatic accumulation from active liver targeting.

Fig. 4.

CS/CNA improves oral absorption and preferential hepatic accumulation after oral administration (A) In vivo fluorescence images of mice after oral administration of free Cy5 or Cy5-CS/CNA at different time points; (B) Ex vivo fluorescence images of isolated intestines after oral administration of free Cy5 or Cy5-CS/CNA; (C) Schematic illustration of the intestinal segments used for fluorescence quantification; (D–F) Quantitative fluorescence intensity in the duodenum, jejunum, and ileum after oral administration of free Cy5 or Cy5-CS/CNA; (G) Fluorescence intensity distribution of Cy5-CS/CNA in different intestinal segments at different time points. (H) Ex vivo fluorescence images of major organs, including the heart, liver, spleen, lung, and kidney, after oral administration of free Cy5 or Cy5-CS/CNA; (I) Quantitative analysis of liver fluorescence intensity at 12 h after oral administration; (J) Schematic illustration of the pharmacokinetic study of free CEL and CS/CNA in SD rats; (K) Plasma concentration–time profiles of CEL after oral administration of free CEL or CS/CNA; (L–N) Pharmacokinetic analysis of AUC_(0-t), Tmax, and Cmax after oral administration of free CEL or CS/CNA. The pseudo-color scale bars indicate fluorescence radiant efficiency, with the minimum and maximum values shown for each panel. Data are presented as mean ± SEM; n = 3 for intestinal fluorescence quantification and pharmacokinetic analysis, n = 5 for liver fluorescence quantification. Statistical significance was analyzed using two-tailed unpaired t-test. *p < 0.05, **p < 0.01, ***p < 0.001. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

To further evaluate the oral absorption and systemic exposure of CEL delivered by CS/CNA, pharmacokinetic studies were conducted on Sprague Dawley (SD) rats (Fig. 4J, Supplementary S7A, B). The average pharmacokinetic curves are presented in Fig. 4K. Pharmacokinetic parameters are shown in Table 1 and Supplementary Table S2. The observed AUC_(0-t) of CS/CNA was 17.92 mg h/L, which was 3.12-fold higher than that of free CEL (5.74 mg h/L) (Fig. 4L). Additionally, the maximum serum concentration (C_max) for CS/CNA was 1.403 mg/L, significantly higher than that of CEL (0.31 mg/L), while the time to reach C_max (T_max) for CS/CNA was 0.278 h, significantly lower than the 1.560 h observed for CEL (Fig. 4M and N).

Table 1.

Pharmacokinetic parameters.

		CEL		CS/CNA
		Mean	SD	Mean	SD
AUC(0-t)	mg·h/L	5.739	0.489	17.923	1.852
AUMC(0-t)	mg·h2/L	63.440	6.041	211.109	31.067
MRT(0-t)	h	11.086	1.013	11.760	1.063
VRT(0-t)	h2	58.534	2.725	63.196	8.651
C_last	mg/L	0.178	0.027	0.546	0.211
Tmax	h	1.560	0.439	0.278	0.136
Cmax	mg/L	0.305	0.024	1.403	0.226

These results demonstrated that CS/CNA improved oral absorption and showed preferential hepatic accumulation after oral administration.

3.4. Mechanisms of CS/CNA cellular uptake

We established a Caco-2 monolayer cell model, which demonstrated transport characteristics and metabolic enzymes similar to those found in small intestinal brush border epithelium (Fig. 5A, Supplementary S8A). We measured the permeability of Cy5-CS/CNA and free Cy5 in this Caco-2 monolayer model, revealing that the permeability of Cy5-CS/CNA was significantly greater than that of free Cy5, with a time-dependent increase (Fig. 5B). Given that molecular transport in intestinal epithelial cells mainly occurs through paracellular and transcellular pathways, we monitored the trans-epithelial electrical resistance (TEER) of the Caco-2 monolayer to determine if CS/CNA disrupt tight junctions. Results indicated that TEER values remained stable at 550 Ω cm² during the 12-h transport period for both Cy5-CS/CNA and free Cy5, showing no significant decrease (Supplementary S8B). Moreover, Western blot results showed that CS/CNA treatment did not alter ZO-1 expression in the Caco-2 monolayer model, suggesting that the primary absorption mechanism for CS/CNA in intestinal epithelial cells is likely via transcellular pathways (Fig. 5C and D). The internalization effect of CS/CNA in Caco-2 cells was investigated using flow cytometry and confocal microscopy. Flow cytometry results showed that Caco-2 cells exhibited the highest uptake of Cy5-CS/CNA at 6 h, significantly surpassing the free Cy5 group (Fig. 5E and F). Confocal fluorescence images confirmed this finding, with red fluorescence in Caco-2 cells from the Cy5-CS/CNA group clearly higher than in the free Cy5 group (Fig. 5G and H).

Fig. 5.

Mechanisms of CS/CNA Cellular Uptake. (A) Schematic illustration of the Caco-2 Transwell monolayer model used to evaluate epithelial transport of free Cy5 and Cy5-CS/CNA; (B) Time-dependent transport efficiency of free Cy5 and Cy5-CS/CNA across the Caco-2 monolayer; (C, D) Representative Western blot bands and quantitative analysis of ZO-1 expression in Caco-2 monolayers after treatment with control medium, CS/CNA, or CEL; (E, F) Flow cytometry analysis and corresponding quantification of cellular uptake of free Cy5 and Cy5-CS/CNA in Caco-2 cells at different time points; (G, H) Confocal laser scanning microscopy (CLSM) images and quantitative fluorescence analysis of free Cy5 and Cy5-CS/CNA uptake in Caco-2 cells. Blue: DAPI-stained nuclei; red: Cy5. Scale bar: 50 μm; (I) Flow cytometry analysis of Cy5-CS/CNA uptake in Caco-2 cells after incubation at 4°C or treatment with pharmacological inhibitors commonly used to probe endocytic pathways; (J) Flow cytometry analysis of Cy5-CS/CNA exocytosis/efflux in Caco-2 cells after treatment with pharmacological inhibitors commonly used to probe intracellular trafficking and exocytosis pathways; (K–M) Co-localization analysis of Cy5-CS/CNA with the endoplasmic reticulum (K), lysosomes (L), and Golgi apparatus (M) in Caco-2 cells. Blue: DAPI-stained nuclei; red: Cy5-CS/CNA; green: organelle markers. Scale bars: 25 μm. Pearson's correlation coefficient (R) and thresholded Manders' overlap coefficients (tM1 and tM2) were calculated using ImageJ/Fiji. Data are presented as mean ± SEM, n = 3. Statistical significance was analyzed using two-tailed unpaired t-test or one-way ANOVA, as appropriate. *p < 0.05, **p < 0.01, ***p < 0.001; ns, not significant. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

To further elucidate the internalization mechanism, we placed Caco-2 cells at 4°C to examine whether CS/CNA uptake is energy-dependent. Results demonstrated that 4°C significantly inhibited CS/CNA uptake (inhibition rate of 82.31%), indicating an energy-dependent process (Fig. 5I). We then pre-treated the cells with chlorpromazine (CPZ), amiloride, and methyl-β-cyclodextrin (MβCD), pharmacological inhibitors commonly used to probe clathrin-mediated endocytosis, macropinocytosis, and caveolae-associated pathways, respectively. Treatment with MβCD slightly inhibited CS/CNA uptake (inhibition rate of 8.79%), but co-incubation with CPZ and amiloride significantly reduced uptake (inhibition rates of 47.29% and 61.91%, respectively) (Fig. 5I). These results suggest that clathrin-mediated endocytosis and macropinocytosis may be involved in CS/CNA internalization, with a relatively minor contribution from caveolae-mediated endocytosis. As the endoplasmic reticulum (ER) and Golgi apparatus are key organelles for exocytosis, while lysosomes are the main organelles for degradation, we subsequently treated the cells with Brefeldin A, Monensin, Nocodazole, and Bafilomycin A1—known inhibitors of the ER/Golgi pathway, Golgi/cell membrane pathway, microtubule pathway, and lysosomal pathway, respectively—to further clarify the exocytosis pathway of CS/CNA in Caco-2 cells. Results showed that Brefeldin A, Monensin, and Nocodazole significantly inhibited the exocytosis of CS/CNA, while co-incubation with Bafilomycin A1 did not significantly alter CS/CNA exocytosis (Fig. 5J). Confocal images indicated that after internalization, CS/CNA showed partial co-localization with lysosomes, the ER, and the Golgi apparatus (Fig. 5K, L, M). Quantitative co-localization analysis further supported partial spatial overlap between Cy5-CS/CNA and lysosomes, ER, or Golgi apparatus. These results suggest that ER/Golgi-associated trafficking, Golgi/cell membrane transport, and microtubule-related pathways may participate in the intracellular transport and exocytosis of CS/CNA, together with possible lysosomal escape.

3.5. CS/CNA exhibits favorable biocompatibility in vitro and in vivo

The biocompatibility of CS/CNA was evaluated both in vitro and in vivo. CCK-8 assays showed that CS/CNA did not markedly affect the viability of Caco-2 and HepG2 cells after 24 h incubation, with cell viability remaining above 85% across the tested concentrations (Supplementary Fig. S9). In contrast, free CEL significantly reduced HepG2 cell viability in a concentration-dependent manner (Supplementary Fig. S9B), indicating that CS/CNA reduced CEL-associated cytotoxicity.

The short-term safety of CS/CNA was further evaluated after 7-day repeated oral administration. Mice treated with high-dose free CEL showed significant body weight loss, whereas no obvious body weight reduction was observed in the CS/CNA-treated groups (Supplementary Fig. S10). In addition, high-dose CEL markedly increased serum ALT, AST, UREA, and CK levels and induced mild histopathological damage in the liver and spleen. By contrast, CS/CNA treatment did not cause obvious abnormalities in hematological or serum biochemical parameters, nor did it induce apparent histopathological damage in major organs, including the heart, liver, spleen, lung, kidney, and brain (Fig. 6B–D).

Fig. 6.

Short-term safety evaluation of CS/CNA after 7-day repeated oral administration in mice. (A) Schematic illustration of the 7-day oral safety evaluation. Mice were orally administered vehicle, low-dose CEL, low-dose CS/CNA, high-dose CEL, or high-dose CS/CNA once daily for 7 consecutive days. Low and high doses correspond to 4 and 20 mg/kg CEL equivalent, respectively; (B) Serum biochemical parameters, including ALT, AST, TBIL, CREA, LDH, CK, and UREA; (C) Hematological parameters, including RBC, MCV, PLT, MCH, MCHC, and HGB; (D) Representative H&E-stained images of major organs, including the heart, liver, spleen, lung, kidney, and brain. Scale bars: 100 μm. Data are presented as mean ± SEM, n = 4. Statistical significance was analyzed by one-way ANOVA. **p < 0.01, ***p < 0.001.

Considering that NASH requires long-term treatment, we further performed a 30-day repeated-dose safety study using low-, medium-, and high-dose CS/CNA, corresponding to 4, 10, and 20 mg/kg CEL equivalent, respectively (Supplementary S11A). During the treatment period, no mortality, abnormal behavior, or obvious body weight loss was observed (Supplementary S11C). Gross organ observation, organ coefficient analysis, serum biochemical analysis, hematological examination, and H&E staining revealed no apparent treatment-related abnormalities after 30-day CS/CNA administration (Supplementary S11B-F and Supplementary S12).

Overall, these results demonstrate that CS/CNA has favorable cellular compatibility, reduces CEL-associated toxicity, and exhibits good repeated-dose safety in mice.

3.6. Dose-gradient evaluation and therapeutic efficacy of CS/CNA in non-alcoholic steatohepatitis (NASH) mice

We first performed a dose-gradient study to determine the appropriate therapeutic dose of CS/CNA in HFMCD-induced NASH mice. Mice were fed an HFMCD diet for 8 weeks and orally administered low-, medium-, or high-dose CS/CNA during the last 4 weeks, corresponding to 4, 10, or 20 mg/kg CEL equivalent, respectively (Supplementary S13A). Compared with the HFMCD group, CS/CNA treatment improved serum liver injury markers, gross liver morphology, liver weight/body weight ratio, hepatic steatosis, fibrosis, and lipid accumulation in a dose-related manner (Supplementary S13B-J). Among the tested doses, medium-dose CS/CNA achieved substantial therapeutic improvement, whereas high-dose CS/CNA did not show markedly superior effects in several endpoints. Therefore, medium-dose CS/CNA was selected as the representative dose for subsequent detailed pharmacodynamic evaluation.

Based on the dose-gradient results, medium-dose CS/CNA was further evaluated in the HFMCD-induced NASH model, with Resmetirom used as a positive control. Mice were divided into five groups: NC, HFMCD, HFMCD-CEL, HFMCD-CS/CNA, and HFMCD-Resmetirom. From week 4 to week 8, mice were orally administered vehicle, free CEL, medium-dose CS/CNA, or Resmetirom daily (Fig. 7A). Compared with the HFMCD and CEL groups, CS/CNA and Resmetirom significantly reduced serum ALT, AST, and TC levels (Fig. 7B, C, E). GTT and ITT results showed that HFMCD feeding induced glucose intolerance and insulin resistance, whereas CS/CNA and Resmetirom treatment improved glucose tolerance and insulin sensitivity (Fig. 7F–J). Consistently, CS/CNA and Resmetirom improved gross liver morphology and reduced the liver weight/body weight ratio (Fig. 7K). Histological analysis further showed that CS/CNA and Resmetirom alleviated hepatic steatosis, hepatocyte ballooning, inflammatory infiltration, fibrosis, and lipid accumulation, as evidenced by H&E, Masson, Sirius Red, and Oil Red O staining (Fig. 7L–O). Western blot analysis also showed decreased expression of IL-6, TNF-α, and ACTA2 in the CS/CNA and Resmetirom groups compared with the HFMCD and CEL groups (Supplementary Fig. S14).

Fig. 7.

CS/CNA Exhibit Therapeutic Advantages in the Treatment of Non-Alcoholic Steatohepatitis (NASH). (A) Schematic illustration of the HFMCD-induced NASH model and oral administration schedule. Mice were fed an HFMCD diet for 8 weeks and orally administered vehicle, free CEL, Resmetirom, or CS/CNA during weeks 4–8; (B–E) Serum levels of ALT, AST, TG, and TC in different treatment groups; (F) Schematic illustration of the glucose tolerance test (GTT) and insulin tolerance test (ITT); (G, I) Blood glucose curves and corresponding area under the curve (AUC) analysis during GTT; (H, J) Blood glucose curves and corresponding AUC analysis during ITT; (K) Representative gross liver images and quantitative analysis of liver weight/body weight ratio. Scale bar: 1 cm; (L) Representative H&E-stained liver sections and NAS score analysis. Scale bar: 100 μm; (M) Representative Masson-stained liver sections and quantitative analysis of fibrotic area. Scale bar: 100 μm; (N) Representative Sirius Red-stained liver sections and quantitative analysis of Sirius Red-positive area. Scale bar: 250 μm; (O) Representative Oil Red O-stained liver sections and quantitative analysis of Oil Red O-positive area. Scale bar: 100 μm. Data are presented as mean ± SEM, n = 5. Statistical significance was analyzed by one-way ANOVA. *p < 0.05, **p < 0.01, ***p < 0.001. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

These findings suggest that CS/CNA exerts promising anti-inflammatory, anti-fibrotic, and lipid-lowering effects in the treatment of NASH.

4. Discussion

Celastrol has attracted considerable interest for NASH intervention because of its lipid-modulating, anti-inflammatory, and hepatoprotective activities [[13], [14], [15], [16], [17], [18], [19]]. However, its therapeutic application is limited by poor aqueous solubility, insufficient oral exposure, and dose-limiting toxicity [13,[20], [21], [22]]. Therefore, the key formulation challenge is not only to increase the apparent solubility and intestinal absorption of CEL, but also to reduce direct systemic exposure to free CEL before it reaches the therapeutic site. In this context, the CS/CNA system was designed as a natural polysaccharide-guided, carrier-free supramolecular nanoassembly to integrate drug dispersion, gastrointestinal stability, epithelial transport, and toxicity reduction within a single oral formulation strategy.

The selection of CS was central to this formulation logic. In the initial screening, CS/CNA showed a relatively uniform nanoscale size, low PDI, moderately negative surface charge, and well-dispersed spherical morphology compared with other polysaccharide-CEL nanoassemblies. These features are important for oral nanoformulation development because they can influence colloidal stability, aggregation tendency, mucus/barrier interactions, and epithelial transport behavior [[41], [42], [43]]. The additional hemolysis assay further expanded the screening framework beyond physicochemical characterization and provided an initial hemocompatibility comparison among representative polysaccharide-CEL nanoassemblies. Nevertheless, this screening strategy was still mainly based on formulation developability and in vitro compatibility. Systematic comparisons of enzymatic degradation, biodistribution, tissue pharmacokinetics, and in vivo metabolic fate among all polysaccharide-CEL nanoassemblies were not performed. This limitation should be considered when interpreting the selection of CS as the optimal polysaccharide companion, and these biological and metabolic dimensions could be incorporated into future screening frameworks for polysaccharide-based oral nanoassemblies.

The structural characterization provides a mechanistic basis for the improved oral performance of CS/CNA. XRD analysis indicated that the crystalline peaks of CEL disappeared after nanoassembly formation, suggesting that CEL was transformed into an amorphous or molecularly dispersed state within CS/CNA. This structural transition may facilitate CEL dispersion in aqueous gastrointestinal environments and contribute to the improved observed systemic exposure after oral administration [44]. In addition, the stability of CS/CNA in SGF, SIF, different pH conditions, and salt-containing media suggests that the nanoassembly can maintain its colloidal structure under physiologically relevant conditions. This stability is important because premature disassembly or aggregation in the gastrointestinal tract would reduce the probability of intact nanoparticle transport across the intestinal barrier [42,45].

The cellular transport results further support the oral absorption logic of CS/CNA. In the Caco-2 monolayer model, CS/CNA showed enhanced permeability without obvious disruption of epithelial integrity, indicating that the improved transport was unlikely to be primarily caused by tight-junction damage. Instead, the inhibitor-based results suggested that energy-dependent endocytosis, particularly clathrin-mediated endocytosis and macropinocytosis, may contribute to CS/CNA internalization. Intracellular trafficking and co-localization analyses further suggested possible involvement of ER/Golgi-associated transport, microtubule-related trafficking, and partial lysosomal escape. These processes may help preserve nanoassembly integrity during epithelial transport and thereby contribute to the higher plasma exposure observed for CS/CNA. However, because pharmacological inhibitors may have limited specificity and can perturb multiple cellular processes, these pathway assignments should be interpreted cautiously. These results indicate possible transport routes rather than definitive evidence of receptor-mediated uptake.

The therapeutic benefit of CS/CNA appears to arise from the combined improvement in CEL exposure and tolerability. Free CEL is constrained by a narrow therapeutic window, in which effective dosing is limited by systemic toxicity [13,[20], [21], [22]]. By incorporating CEL into the CS-guided nanoassembly, CS/CNA reduced CEL-associated cytotoxicity and attenuated the body weight loss, serum biochemical abnormalities, and histopathological damage observed after repeated free CEL administration. This improved safety profile is important for NASH treatment, as chronic liver diseases generally require repeated or long-term pharmacological intervention. In the HFMCD-induced NASH model, CS/CNA improved several disease-relevant endpoints, including liver injury, glucose and insulin tolerance, hepatic steatosis, inflammatory responses, and fibrosis-related pathological changes. The comparison with resmetirom further supports the therapeutic potential of CS/CNA; however, these findings should be interpreted as comparable improvements in several preclinical endpoints rather than direct clinical equivalence. Additional NASH models and longer treatment periods would be required to further evaluate the robustness and durability of the therapeutic response.

Several limitations should be considered when interpreting these findings. First, the MD simulation used a simplified CS tetrasaccharide model, which cannot fully capture the chain length, sulfation heterogeneity, and structural complexity of native CS. Although interaction-disruption assays and FTIR analysis provided experimental support for the involvement of hydrogen bonding and hydrophobic interactions, a more complete understanding of the CS–CEL assembly mechanism would require more representative CS models and complementary biophysical analyses, such as NMR and ITC. Second, although CS/CNA showed preferential hepatic accumulation after oral administration, the current data cannot distinguish passive first-pass hepatic accumulation from possible receptor-mediated processes. This issue could be further clarified by free CS competition, receptor-blocking, or receptor knockdown experiments in intestinal epithelial cells and hepatocytes. Third, the prolonged intestinal fluorescence retention of Cy5-CS/CNA suggests improved intestinal residence, but direct mucus diffusion or mucin-binding assays were not performed. Therefore, the potential mucus interaction or mucus transport behavior of CS/CNA should be interpreted cautiously. Finally, fluorescence imaging provides only qualitative or semi-quantitative biodistribution information. Quantitative LC-MS/MS-based tissue pharmacokinetic analysis would be needed to determine CEL accumulation, metabolite formation, hepatic clearance, and off-target tissue exposure. In addition, scalable preparation, batch-to-batch reproducibility, storage stability, and quality-control parameters remain important issues for further translational development.

In conclusion, this study presents a natural polysaccharide-guided supramolecular nanoassembly strategy for improving the oral delivery of poorly soluble drugs. CS/CNA enhanced the oral absorption and observed systemic exposure of CEL, reduced CEL-associated toxicity, and showed promising therapeutic efficacy in HFMCD-induced NASH mice. Although further studies are required to clarify the detailed self-assembly mechanism, hepatic accumulation mechanism, long-term safety, and quantitative tissue pharmacokinetics, this work provides a promising platform for developing safer and more effective oral CEL-based therapeutics for NASH intervention.

CRediT authorship contribution statement

Liming Yang: Conceptualization, Data curation, Formal analysis, Investigation, Software, Validation, Visualization, Writing – original draft. Yunfei Zhao: Conceptualization, Data curation, Methodology, Resources, Software, Writing – review & editing. Jianguo Luo: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Validation. Mengjiao Deng: Formal analysis, Investigation, Methodology, Validation. Fangliang Wang: Formal analysis, Investigation, Methodology, Validation. Gang Zou: Formal analysis, Investigation, Methodology, Validation. Jun Chen: Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Validation. Yunyu He: Methodology, Software, Validation. Xuelan Gan: Visualization. Yumei Dai: Visualization. Yuminghang Chen: Visualization. Chao Yu: Conceptualization, Funding acquisition, Project administration, Resources, Supervision, Writing – review & editing.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following is the Supplementary data to this article.

Data availability

Data will be made available on request.