Metabolomics-driven elucidation of the synergistic therapeutic mechanism of a novel SGLT-2/PPAR-γ dual receptor supramolecular system for treatment diabetes and obesity

Abstract

A supramolecular system of active pharmaceutical ingredients (APIs) can modify the physicochemical properties and enhance the synergistic efficacy of their components; however, the relevant underlying mechanisms in vivo remain unclear. This study employed a metabolomics-driven approach, combined with biological validation, to investigate the synergistic mechanisms of API-based supramolecular systems. Metabolic dysfunction exacerbates insulin resistance and obesity, contributing to hepatic steatosis and cardiac hypertrophy. A novel sodium-dependent glucose transporter 2 (SGLT-2)/peroxisome proliferator-activated receptor-γ (PPAR-γ) dual receptor (dapagliflozin-pioglitazone (DAP-PIO)) supramolecular system was selected as the model to explore the synergistic mechanism involved in the treatment of metabolic dysfunctions, diabetes and obesity. First, metabolomics analyses were performed to compare the effects of a simple physical mixture (PM) of DAP and PIO with the DAP-PIO supramolecular system after absorption into the bloodstream. The results demonstrated significant differences, with the supramolecular system activating the phosphatidylinositol 3-kinase (PI3K)/protein kinase B (AKT) and adenosine monophosphate-activated protein kinase (AMPK) signaling pathways. Ceramide (Cer), a key metabolite in sphingolipid metabolism, emerged as a critical mediator. Subsequently, the mechanisms underlying the DAP-PIO supramolecular system’s hypoglycemic effects and its ability to ameliorate hepatic steatosis and myocardial hypertrophy by reducing insulin resistance were evaluated and confirmed. These findings provide an innovative strategy for developing SGLT-2/PPAR-γ dual-receptor supramolecular systems to enhance the therapeutic outcomes for diabetes and obesity.

Graphical abstract

Highlights

• •A novel SGLT-2/PPAR-γ dual receptor supramolecular system was prepared.
• •Metabolomics was used to elucidate the synergistic therapeutic mechanism in diabetes and obesity.
• •The supramolecular activated PI3K/AKT and AMPK pathways featured by sphingolipid metabolism.
• •The insulin resistance was relieved effectively by supramolecular in the treatment of diabetes and obesity.

1. Introduction

Approximately 40% of the candidate active pharmaceutical ingredients (APIs) in the pharmaceutical industry are only slightly soluble in water, which poses a significant challenge in improving their oral bioavailability [1]. Addressing this issue has been a critical focus in the field of drug development. Supramolecular systems, such as co-crystals (COC) and co-amorphous (COA) phases, are formed through the non-covalent interactions of two components, including hydrogen bonds. These systems have the potential to enhance drug solubility, dissolution rates, and bioavailability, thereby representing a promising strategy for pharmaceutical innovation [2]. However, it remains unclear whether the pharmacological effects of supramolecular systems arise from functioning as a “physical mixture (PM) of API and co-former (CCF)” or as a “new entity based on weak force” [3]. Recent studies have demonstrated that supramolecular systems exhibit distinct biological behaviors compared to PMs [4,5]. Nonetheless, how the different forms of these components in PMs and supramolecular systems influence the in vivo and in vitro behaviors and, subsequently, their therapeutic efficacy remains poorly understood.

Metabolomics, a rapidly advancing analytical technology, has been increasingly applied across various fields [6]. In a previous study, cell metabolomic was employed to investigate the cytotoxic mechanisms of a 5-fluorouracil-phenylalanine COC. This research revealed differences in purine and glycerophospholipid metabolism between the PM and COC. Specifically, downregulated deoxyguanosine diphosphate and adenosine diphosphate in purine metabolism, along with downregulated l-glycerophosphocholine and upregulated C16-dihydroceramide in glycerophospholipid metabolism, were linked to cellular antiproliferation and apoptosis, which directly impacted cytotoxicity. These findings highlighted the potential of metabolomics as a powerful and innovative tool for elucidating the synergistic mechanisms of two-component supramolecular systems [7]. However, when supramolecular systems are absorbed into the bloodstream via oral administration and cross the intestinal barrier, it is essential to determine whether they retain their supramolecular advantages and exhibit distinct synergistic mechanisms. Further investigation is required to understand whether supramolecular systems differ metabolically from PMs in vivo, particularly in the context of diseases associated with metabolic dysfunction.

Diabetes and obesity are hallmark metabolic dysfunction disorders [8]. Metabolic dysfunction exacerbates insulin resistance and obesity, which contributes to complications such as hepatic steatosis and cardiac hypertrophy. Dapagliflozin (DAP), a selective sodium-dependent glucose transporter 2 (SGLT-2) inhibitor derived from the natural product phlorizin, is widely used in diabetes treatment. However, DAP faces challenges such as hygroscopicity and the risk of phase transition [9]. As a phlorizin analog, DAP exerts hypoglycemic effects by continuously and stably controlling blood glucose (GLU) levels independent of insulin secretion. This mechanism preserves β-cell function by preventing overstimulation, which could otherwise lead to drug resistance and β-cell decline. Beyond glycemic control, DAP also improves insulin sensitivity, reduces body weight, mitigates the risk of adverse renal and cardiovascular events, and prolongs survival in patients with chronic kidney proteinuria [10]. Research by Yang et al. [11] demonstrated that DAP significantly reduces the total body fat, subcutaneous fat, and visceral adipose tissues, making it highly effective in the early and middle stages of type 2 diabetes mellitus (T2DM). However, DAP is associated with several side effects, including hypotension, ketoacidosis, acute kidney injury, renal impairment, urosepsis, pyelonephritis, genital fungal infections, increased low-density lipoprotein cholesterol (LDL-C), and bladder cancer. Moreover, its hypoglycemic efficacy diminishes in patients with declining renal function, particularly in advanced T2DM [12]. Combination therapy has emerged as a clinically effective strategy to address these limitations.

Pioglitazone (PIO), a thiazolidinedione-class hypoglycemic drug and peroxisome proliferator-activated receptor-γ (PPAR-γ) receptor agonist, is an insulin-dependent agent commonly used as an insulin sensitizer [13]. The efficacy of PIO decreases with the decline of pancreatic β-cell function during the progression of T2DM. In addition, the weight gain induced by PIO can further exacerbate insulin resistance [14]. Moreover, the long-term administration of PIO has side effects on the liver, kidneys, and heart, including abnormal liver function, edema, mild to moderate anemia, and induced heart failure. Combination therapy is often employed to address these challenges, including secondary drug resistance and side effects associated with PIO. Past studies have indicated that combining DAP and PIO is more effective than monotherapy in reducing glycosylated hemoglobin (HbA1c) levels and body weight [15]. Notably, this combination does not amplify PIO-related side effects. DAP’s mild diuretic properties counteract PIO-induced weight gain and fluid retention, while its GLU-removal action complements PIO’s insulin-sensitizing effects. Furthermore, this the combination alleviates adverse outcomes such as acute kidney injury and renal impairment caused by DAP. The cardiovascular benefits of DAP mitigate the cardiac and hepatic toxicity associated with PIO. In summary, the dual-target combination therapy of DAP and PIO holds significant promise for achieving long-term, stable blood sugar control, particularly in advanced-stage diabetes, while reducing the risk of adverse effects. However, the complementary mechanisms underlying this combination therapy require further elucidation.

In this study, the DAP-PIO supramolecular systems were initially prepared by using rotary evaporation methods. The resulting supramolecular systems were then characterized through thermogravimetry (TG), powder X-ray diffraction (PXRD), Fourier-transformed infrared (FT-IR) spectroscopy, Raman spectroscopy, proton nuclear magnetic resonance (¹H NMR), and temperature-modulated differential scanning calorimetry (TMDSC). Their physicochemical properties and the in vivo bioavailability were then systematically evaluated. Metabolomics analyses were employed to identify the differences in the metabolites between the PM of DAP/PIO and the DAP-PIO supramolecular system after their absorption into the bloodstream of the diabetes and obesity models. The results revealed that the supramolecular system, in contrast to the PM, activated the phosphatidylinositol 3-kinase (PI3K)/protein kinase B (AKT) and adenosine monophosphate-activated protein kinase (AMPK) signaling pathways (Scheme 1). In addition, the mechanism through which the DAP-PIO supramolecular system exerts hypoglycemic effects was verified to reverse hepatic steatosis, and alleviates myocardial hypertrophy via a reduction in insulin resistance was evaluated and verified through both in vitro and in vivo systems. These findings provide innovative insights and theoretical guidance toward the clinical application of DAP and PIO combination therapy.

Scheme 1.

Schematic diagram of the pharmacodynamic effects and mechanisms of the sodium-dependent glucose transporters 2 (SGLT-2)/peroxisome proliferator-activated receptor-γ (PPAR-γ) dual receptor supramolecular system in the treatment of diabetes and obesity. DAP: dapagliflozin; PIO: pioglitazone; AMPK: adenosine monophosphate-activated protein kinase; GLUT4: glucose transporter 4; PI3K: phosphatidylinositol 3-kinase; AKT: protein kinase B; GSK: glycogen synthase kinase; IR: insulin receptor; IRS: insulin receptor substrate.

2. Materials and methods

2.1. Reagents

DAP (CAS: 461432-26-8) and PIO hydrochloride (CAS: 112529-15-4) were purchased from Zhengzhou Alpha Chemical Co., Ltd. (Zhengzhou, China). Dimethyl sulfoxide (DMSO) and ethanol were acquired from Fuyu Fine Chemical Co., Ltd. (Tianjin, China). 3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium ammonium bromide (MTT) was sourced from Inokai Technology Co., Ltd. (Beijing, China). Dulbecco’s modified Eagle’s medium (DMEM), fetal bovine serum (FBS), pen-strep double antibody (PS), and 0.25% trypsin (containing 0.02% ethylenediaminetetraacetic acid (EDTA)) were purchased from Biological Industries (Beit Haemek, Northern District, Israel). LY294002 (CAS: 154447-36-6), dorsomorphin (compound C) dihydrochloride (CAS: 1219168-18-9), and 2-deoxy-2-[(7-nitro-2,1,3-benzoxadiazol-4-yl) amino]-d- glucose (2-NBDG) (186689-0706) were procured from Cayman Chemical Company (Ann Arbor, MI, USA). A GLU test kit was bought from Nanjing Jiancheng Bioengineering Institute (Nanjing, China). Glucose transporter 4 (GLUT4) rabbit polyclonal antibody was sourced from Shanghai Beyotime Biotechnology Co., Ltd. (Shanghai, China). Other chemicals, such as total cholesterol (TC) content determination kit, triglycerides (TRIG) content determination kit, high-density lipoprotein cholesterol (HDL-C) determination kit, LDL-C determination kit, alanine aminotransferase (ALT) determination kit, and aspartate aminotransferase (AST) determination kit were acquired from Jiangsu Jingmei Biotechnology Co., Ltd. (Yancheng, Jiangsu, China).

2.2. Preparation of the DAP-PIO supramolecular system

Rotary evaporation method: 200 mg of the PMs containing DAP and PIO with molar ratios of DAP/PIO 1:2, 1:1, 2:1 was placed into a 250 mL of round-bottom flask, followed by adding 140 mL of ethanol. After dissolving completely under ultrasonication at room temperature for 30 min, the solution was evaporated at 38 °C for 20 min at a rotation speed of 50 rpm to remove the solvent. Then, the product was dried overnight in a 40 °C vacuum oven.

Detailed structural characterization and physicochemical property evaluation methods are described in Sections 1.2 and 1.3 of the Supplementary data.

2.3. Pharmacokinetic study

Briefly, 24 male Sprague-Dawley rats weighing 250 ± 25 g (Experimental Animal Center of Hebei Province, SPF (Beijing) Biotechnology Co., Ltd., Beijing, China, License No.: SCXK (Ji) 2022-001) were used to conduct the pharmacokinetic study. All animal handling protocols adhered to the regulations for the care and use of laboratory animals and were approved by the Animal Experimentation Ethics Committee of Hebei Medical University (Approval No.: IACUC-Hebmu-2023036). The Sprague-Dawley rats were fed ad libitum with standard laboratory food and water and subsequently fasted for 12 h before administering the experimental doses. The rats were randomly assigned to the following four groups, with six rats in each group, and subjected to the following intervention: DAP group (83 mg/kg of DAP), PIO group (80 mg/kg of PIO), PM group (a combination of 83 mg/kg DAP and 80 mg/kg PIO), and COA-I group (comprising components of 83 mg/kg DAP and 80 mg/kg PIO). Pharmacokinetic analyses were performed to observe the drug dynamics in the body using safe, elevated doses, ensuring no irreversible harm to the experimental animals [16]. Post-dosing, 0.3 mL blood samples were collected from the orbital sinus at 5, 15, and 30 min, and 1, 1.5, 2, 3, 5, 8, 10, 16, and 24 h into heparinized tubes, then centrifuged at 3,500 rpm for 10 min at 4 °C. The complete methods for plasma sample processing and pharmacokinetic analysis are detailed in Section 1.4 of the Supplementary data.

2.4. Metabolomics study

C57BL/6J mice (Male, age: 4 weeks) were procured from SPF (Beijing) Biotechnology Co., Ltd. (License No.: SCXK (Jing) 2024-0001) and subjected to a 1-week acclimatization period. All animal handling protocols adhered strictly to the regulations for the care and use of laboratory animals and received approval from the Animal Experimentation Ethics Committee of Hebei Medical University (Approval No.: IACUC-Hebmu-2023036). A normal control group was maintained on a standard diet, whereas the diabetic model group was administered a high-fat diet (HFD) comprising 60% energy from fat (D12492, SPF (Beijing) Biotechnology Co., Ltd.). The feed was stored at −4 to −15 °C to preserve its freshness and hardness. After fed with HFD for 2.5 months, the mice were intraperitoneally injected with streptozotocin (STZ) solution at a dose of 50 mg/kg/d continued for 1 week. Mice with fasting blood glucose (FBG) levels ≥ 11.1 mmol/L were categorized as models for T2DM and continued to be fed with HFD. The control group was fed with a standard diet. The body weight of all mice was recorded weekly. Metabolomics study was performed using normal diet mice and diabetic model mice. The normal diet mice were grouped into control, co-amorphous (C-COA), and the complex (C-PM) groups, while the diabetic model mice were divided to model (MOD), co-amorphous (M-COA), and the complex (M-PM). Each experiment was implemented in 6 groups in parallel (n = 6). The mice received oral gavage of daily at a dose of 2 mg/kg of DAP and 1.92 mg/kg of PIO continued for 3 weeks. Detailed protocols for metabolite extraction, measurement, and data analysis are provided in Section 1.5 of the Supplementary data. “MOD group” is the abbreviation for “model group,” and in our study, it indeed serves as a core concept with specific references depending on the experimental system. In animal-level studies, the “MOD group” specifically refers to the diabetic model mice established through HFD induction (i.e., HFD-WT diabetic mice). This group is used to simulate in vivo metabolic disorder conditions. In cellular-level studies, the “MOD group” refers to cells exposed to high-glucose, high-lipid conditions in vitro. This group is used to simulate the direct impact of the diabetic pathological environment on cells.

2.5. In vitro cell pharmacological efficacy evaluation

2.5.1. Cell culture and treatment

HepG2 and H9c2 cells were cultured in DMEM medium containing 25 mM GLU, supplemented with 10% FBS and 1% penicillin/streptomycin at 37 °C in a 5% CO₂ atmosphere. To simulate T2DM in vitro, a hyperglycemic hyperinsulinemic (HGHI) condition was established by treating cells with 30 mM GLU and 5 μM insulin for 30 h. Cells under HGHI conditions were treated for 24 h in the absence or presence of different concentrations of DAP, PIO, PM of DAP and PIO, and DAP-PIO COA. The cytotoxic effects of these treatments were subsequently evaluated using the MTT assay. Methods are detailed in Section 1.6 of the Supplementary data.

2.5.2. GLU consumption rate measurement

Cells were cultured under HGHI conditions and treated for 24 h with or without different concentrations of DAP, PIO, PM, and the complex. In the 24 h period, the GLU concentration in the culture medium of each group was measured using a GLU test kit (Nanjing Jiancheng, Nanjing, China). The GLU consumption rate was calculated by the formula: (GLU concentration of each group − GLU concentration of the control group) ÷GLU concentration of the control group × 100%.

2.5.3. GLU uptake assay in insulin-resistant HepG2 and H9c2 cell models

The cells were seeded in 6-well plates with cover slides. After 24 h of incubation under HGHI conditions with or without different drugs, 2-NBDG (fluorescently labeled GLU analogue to detect GLU uptake in cells) was added at 37 °C for 30 min, cells were fixed with 4% paraformaldehyde solution for 20 min, finally nuclei were stained with DAPI for 20 min, and plates were subsequently sealed. The fluorescence intensity of 2-NBDG was observed on a laser scanning con-focal microscopy (CLSM) (con-focal-LSM-900, Zeiss, Oberkochen, Germany) with an excitation at 488 nm and emission at 550 nm. Cells were seeded in 6-well plates and cultured under HGHI conditions with or without different drugs for 24 h. After the incubation, cells were washed twice with PBS, following adding of 2-NBDG with action concentration of 30 μM and stained for 20 min. The cells were washed twice with PBS after removing the fluorescent dye, digested with trypsin, then quenched with 1 mL of PBS, and collected by centrifugation. The collected cells were washed twice with pre-cooled heparin sodium solution (1 mg/mL) and re-collected by centrifugation, then resuspended in 500 μL of PBS, followed by determination of the uptake of 2-NBDG in the cells using FACSVerse™ flow cytometer (Becton, Dickinson, Franklin Lakes, NJ, USA).

2.5.4. Synergistic mechanism of blood sugar reduction

The intrinsic interactions of DAP, PIO, and assembled molecules with two protein receptors were investigated by molecular docking according to the methods illustrated in Ref. [17]. Detailed methodology is available in Section 1.7 of the Supplementary data.

2.5.5. GLUT4 immunofluorescence

For mechanistic studies, prior to HGHI incubation, cells were pre-treated for 2 h with the PI3K inhibitor LY294002 and AMPK inhibitor compound C. Cells under HGHI conditions were then treated for 24 h with or without different concentrations of DAP, PIO, PM, and COA-I. Thereafter, the cells were incubated with GLUT4 Rabbit Polyclonal Antibody (Shanghai Beyotime Biotechnology Co., Ltd., Shanghai, China), followed by incubation with the corresponding fluorescein-conjugated secondary antibody. After washing with PBS, the cells were observed by CLSM (con-focal-LSM-900, Zeiss).

2.6. The in vivo assay on hyperglycemia and insulin sensitivity

As outlined in Section 2.4, which details the animal modeling methods, the T2DM model mice were allocated into five distinct groups: HFD-WT diabetic mice group (MOD), DAP, PIO, PM, and COA, each consisting of six mice that were continuously fed with HFD. In contrast, the control group comprised six mice that were maintained on a standard diet. The body weight of all mice was recorded weekly. The mice received oral gavage of daily at a dose of 2 mg/kg of DAP and 1.92 mg/kg of PIO continued for 3 weeks.

2.6.1. Oral glucose tolerance test (OGTT) and insulin tolerance test (ITT)

After 3-week oral gavage administration, the FBG levels of the mice were measured using a Roche glucometer. For OGTT and ITT, the mice were fasted overnight, followed by oral administration of 1 g/kg GLU or intraperitoneal injection of 0.75 U/kg insulin. The GLU baseline levels were measured at 0, 15, 30, 60, 90, and 120 min and calculated using the area under the curve (AUC). After euthanizing the mice, their plasma and tissue samples were collected. The body weight, liver weight, and heart weight were recorded and the liver and heart indices were calculated.

2.6.2. Homeostasis model assessment-insulin resistance (HOMA-IR) index

After the drug administration ended, the mice fasted for 12 h and the body weight was recorded. The fasting plasma glucose (FPG) level of each group was tested using a Roche glucometer (Roche Diabetes Care GmbH, Mannheim, Germany). Fasting insulin levels (FINS) were determined using an enzyme-linked immunosorbent assay (ELISA), and insulin resistance was evaluated using the HOMA-IR index, which was calculated as described in Ref. [18].

2.6.3. Liver function and lipid profile tests

After the drug administration ended, three mice were randomly selected from each group, and blood was taken from the orbital vein. Serum was prepared using the orbital blood by centrifugation at 3,000 rpm, and the lipid profile tests for total TC, TRIG, HDL-C, and LDL-C were performed using their respective test kits. Liver function tests (ALT, AST) were also conducted using ALT and AST reagent kits.

2.6.4. Histopathological staining

Liver, heart, pancreas, spleen, and kidney from HFD-induced diabetic mice were collected. Hematoxylin and eosin (H&E), and Oil Red O staining were used to observe the morphological structures of the liver and heart and to display fat within the tissue. Periodic acid-schiff (PAS) staining was utilized to observe glycogen deposition in the liver and heart. Insulin immunoperoxidase staining and immunofluorescence were employed to detect insulin localization and expression, with images captured using an optical microscope.

2.6.5. RNA extraction and quantitative real-time polymerase chain reaction (qRT-PCR) analysis

The messenger RNA (mRNA) in liver and heart tissues was extracted using the Trizol reagent kit (R6934-01, Omega Bio-Tek, Norcross, GA, USA) and quantified using a nano-spectrophotometer (ND-100, Hangzhou Milan Instrument Co., Ltd., Hangzhou, China). Complementary DNA (cDNA) was synthesized by reverse transcription following the instructions provided with the kit. The expression of target genes was measured by qRT-PCR using a universal SYBR Green Supermix (220629, Mona Biotechnology Co., Ltd., Suzhou, China) on a Real-Time PCR system (7500, Thermo Fisher Scientific, Waltham, MA, USA). Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as the reference gene, and the data were processed by the 2^-△△Ct method. The gene-specific primer sequences used are listed in Table S1.

2.6.6. Western blotting

After processing liver and heart tissues, total protein within the tissues was lysed in radio immunoprecipitation assay (RIPA) lysis buffer. The lysates were then separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), transferred to polyvinylidene fluoride (PVDF) membranes, and incubated with primary antibodies against phosphorylated PI3K (p-PI3K), phosphorylated AKT (p-AKT), and phosphorylated AMPK (p-AMPK). Membranes were probed with specific secondary antibodies, and protein bands were analyzed.

2.7. Statistical analysis

The Student’s t-test was used to compare differences between two groups, and one-way analysis of variance was employed when analyzing differences between three or more groups. Results were considered statistically significant at P < 0.05.

3. Results

3.1. Characterizations

The as-prepared supramolecular complexes were characterized using PXRD, TMDSC, TG, FT-IR, Raman spectroscopy, and ¹H NMR spectra, and the results are presented in Fig. 1. As can be seen in Fig. 1A, crystalline PIO exhibited distinct characteristic diffraction peaks, while the PXRD patterns of the DAP-PIO COA at different ratios display a hump-shaped halo without any sharp characteristic peaks. The disappearance of crystal diffraction peaks preliminarily indicates the formation of a COA state. Here, we marked the samples as COA-I (DAP:PIO = 1:1), COA-II (DAP:PIO = 2:1), and COA-III (DAP:PIO = 1:2), with their corresponding PMs being PM-I (DAP:PIO = 1:1), PM-II (DAP:PIO = 2:1), and PM-III (DAP:PIO = 1:2).

Fig. 1.

Preparation and characterization of dapagliflozin-pioglitazone (DAP-PIO) co-amorphous (COA) phases. (A) Powder X-ray diffraction (PXRD) patterns. (B) Temperature-modulated differential scanning calorimetry (TMDSC) thermograms. (C–F) Thermogravimetry (TG) and derivative thermogravimetry (DTG) curves of the active pharmaceutical ingredient (API) (C), COA-I (D), COA-II (E), and COA-III (F), displaying their thermal decomposition behavior. (G) Fourier-transformed infrared (FT-IR) spectra. (H) Raman spectra. (I) Proton nuclear magnetic resonance (¹H NMR) spectra. PM: physical mixture; T_g: glass transition temperature.

Generally, the formation of the COA phase can lead to the disappearance of the melting points of the respective components and the emergence of a single glass transition temperature (T_g). The T_g is the single most defining characteristic of COA systems. To confirm the formation of COA, not the PM of amorphous state of two components, the TMDSC was used to define the single T_g, the results were presented in Fig. 1B. As shown, the reversible heat flow signal curves of three COAs samples exhibited an endothermic step corresponding to the single T_g. The typical halo patterns in PXRD and the existence of single T_g in TMDSC confirmed the formation of COA phase. T_g values of 26.21, 24.77, 42.12, and 19.75 °C were observed for the DAP, COA-I, COA-II and COA-III, respectively, suggesting that the increases in DAP content rose the T_g of COA. In the TG curves of the COAs, the DTG curve of PIO showed one thermal decomposition peak, indicating a one-step decomposition, while DAP showed two thermal decomposition peaks corresponding to a two-step decomposition (Figs. 1C–F). Furthermore, there were differences between the PM and the COAs, suggesting that weak interaction existed between DAP and PIO in the COAs, but not in the PM.

FT-IR and Raman spectra are used to gain insight into possible the molecular level of weak interaction between the two components in COAs. As shown in FT-IR spectra (Fig. 1G), the vibration absorption of C O in PIO at 1,744 and 1,694 cm⁻¹ shifted to 1,750, 1,696, and 1,698 cm⁻¹ in COAs. Meanwhile, the vibrations assigned to the N–H stretching shifted from 3084 cm⁻¹ in PIO [19] to 3,036 and 3,034 cm⁻¹ in COAs, suggesting that C O and N–H in PIO involved in the formation of hydrogen bonds with DAP. Furthermore, the vibrations assigned to the O–H in DAP [20] moved from 3,378 cm⁻¹ to 3,386, 3,388, and 3,422 cm⁻¹ after the formation of COAs, indicating that the O–H in DAP was involved in the formation of hydrogen bonds.

Besides IR spectra, Raman spectroscopy was used to ascertain the structure and weak interaction at molecular level (Fig. 1H). The absorption of C O in PIO appeared at 1,610 cm⁻¹, and the –OH in DAP was at 873 cm⁻¹ [21]. The physical overlapping peaks of DAP and PIO was observed in the Raman spectra of the PMs. It can be seen that the intensity, width, and position of peaks involving the formation of hydrogen bonds changed in COAs. The peaks at 2,934 and 3,067 cm⁻¹ in the spectra of COAs were significantly broadened, indicating the formation of a disordered structure. At the same time, the –OH vibration of the DAP-PIO supramolecule merges and shifts to 827 cm⁻¹. These notable changes and merging suggest that the co-amorphous material may be formed by intermolecular hydrogen bonding between the –OH of DAP and the –C O of PIO.

The distribution of the peaks in the ¹H NMR for PM and COA is shown in Fig. 1I. Peaks at 2.50 and 3.33 ppm were ascribed to the solvent DMSO‑d₆ and its residual water, respectively [22]. In COA solution, significant changes occurred in the chemical shift, and fragmentation pattern of the α-hydrogen atom peaks of C O for PIO. The chemical shifts of α-hydrogen atom of C O in the PIO shifted from 4.39/4.38/4.37 ppm to 4.38/4.37/4.35 ppm in PM-I and 4.38/4.37/4.36 ppm in COA-I, meanwhile, both PM-II and COA-II showed shifts of 4.37/4.36/4.35 ppm; PM-III and COA-III both had shifts of 4.38/4.37/4.36 ppm. Additionally, the chemical shifts of –NH in the PIO were at 12.01 ppm, and after mixing with DAP, the –NH shifts in both PM and COA became 12.00 ppm. In the DAP drug substance spectrum, the chemical shift at –CH₃ position, 1.31/1.29/1.28 ppm [20], changed to 1.30/1.29/1.28 ppm after mixing, Furthermore, the characteristic peaks at 1.28 ppm and 1.76 ppm in the DAP spectrum disappeared after the formation of the supramolecular system, while they were present in both the parent components of DAP and PM, indicating the presence of intermolecular interactions between DAP and PIO in COA solution. These intermolecular interaction forms are of great significance to the pharmacological activity of the COA. In our previous study of gallic acid cocrystals, we confirmed that hydrogen bonding exists between the interaction of API and CCF molecules in cocrystal solutions through NMR analysis. Molecular docking also showed that the cocrystal binds with α-glucosidase in the form of a supramolecule [17]. Furthermore, increasing reports suggest that cocrystals in solution exhibit different pharmacological activities compared to the PM of API/CCF [23], indicating that when the cocrystal dissolved, API and CCF is not entirely free, but interconnected through intermolecular interactions.

3.2. Physicochemical property evaluation

3.2.1. Equilibrium solubility

The solubility of APIs can affect their bioavailability. The solubility of DAP, PIO, PMs and the COAs in media with pH of 1.0, 1.2, 4.0, and 6.8 were investigated and the results were shown in (Figs. 2A, 2B, S1A, and S1B). The equilibrium solubility of PIO and DAP in COAs showed a decreasing trend as pH increased. Compared to the PMs, the solubility of the COAs were higher with significant differences. After measuring the equilibrium solubility at pH = 1.0, solid samples were collected for PXRD (Figs. 2C and S1C). No diffraction peaks representing crystalline phase appeared in COA-I and II, as well as their PMs, indicating that COA-I and II remained stable without any free base crystallization of PIO after repeated dissolution and precipitation in solvent. However, COA-III and PM-III involving advantageous content of PIO transferred to crystalline state after repeated dissolution and precipitation in solvent.

Fig. 2.

Physicochemical properties of dapagliflozin-pioglitazone (DAP-PIO) co-amorphous (COA)-I. (A, B) Solubility of DAP (A) and PIO (B) in the COA-I system (n = 3). (C) Powder X-ray diffraction (PXRD) at 72 h after pH = 1 (n = 3). (D, E) Dissolution of DAP (D) and PIO (E) in the COA-I system (n = 3). (F) Wetting (n = 3). (G) Stability under high humidity conditions (n = 3). (H, I) Plasma concentration-time profiles of DAP (H) and PIO (I) from COA-I in Sprague–Dawley rats (n = 6). Data are mean ± standard deviations (SD), ^∗P < 0.05 (vs. DAP or PIO), ^∗∗P < 0.01 (vs. DAP or PIO), ^★P < 0.05 (vs. physical mixture (PM)), ^★★P < 0.01 (vs. PM).

3.2.2. Dissolution rates

The dissolution rates of DAP, PIO, PMs and COAs in pH = 1.0 HCl medium were evaluated. As shown in Figs. 2D, 2E, and S1D–G, the dissolution rates increased overall after forming the COAs, especially in COA-I and COA-III. This increase in dissolution rate is beneficial for the bioavailability of the drug in vivo. The dissolution rate of the PMs also improved, which may be due to interactions between DAP and PIO, enhancing each other’s solubility.

3.2.3. Stability assessments

Firstly, the stability of the COAs at high humidity were assayed. The percentage of weight gain after placing the COAs at 25 °C/80% RH for 24 h was calculated Fig. S1H(Figs. 2F and S1H). According to the pharmacopeia’s criteria for hygroscopicity, drugs are considered hygroscopic when the weight gain is less than 15% but higher than 2%. The measured weight gain percentage was 5.36%, 2.57%, and 4.09% for DAP, COA-I, and COA-II samples, respectively, indicating that COAs were less hygroscopic compared to the DAP. The weight gain percentage was 0.86% and 1.61% for PIO and COA-III samples, respectively. According to the pharmacopeia’s criteria, drugs are considered slightly hygroscopic when the weight gain is less than 2% but higher than 0.2%. It indicated that the hygroscopicity of the COAs was inhibited as the ratio of PIO increased. Moreover, samples were subjected to high humidity conditions at 75% RH for 5 and 10 days and the XRD was used to assay the phase stability. As can been seen in Fig. S1IFigs. 2G and S1I. DAP and COA-II remained amorphous. However, COA-I and COA-III showed crystallization after 5 and 10 days, indicating that the stability of the COAs were related with the proportion of PIO. Secondly, the stability of the system was studied at room temperature for 3 weeks and at temperatures exceeding room temperature for 24 h were further assayed. The experimental results revealed that the DAP-PIO supramolecular systems remained stable and amorphous, with no crystallization, after storing at room temperature for 3 week (Fig. S1J) and at temperatures exceeding room temperature for 24 h (Fig. S1K). PXRD analysis confirmed the absence of new diffraction peaks, indicating that the sample is stable and can be stored long-term.

3.2.4. Oral bioavailability

Pharmacokinetic experiments were conducted using the parent components, PM, and the COA following oral administration. Since the clinical dosage of PIO and DAP is typically used at a molar ratio of 1:1, COA-I was selected for the pharmacokinetic assays. The plasma concentration-time curves for DAP and PIO in the rats are presented in Figs. 2H and I, with the corresponding pharmacokinetic parameters summarized in Tables S2 and 3. In our system, the enhanced solubility and dissolution rate of COA-I facilitate an increased bioavailability for DAP once it formed COA-I with PIO. When compared to PM-I and free DAP, the maximum plasma concentration (C_max) of COA-I increased by approximately 1.52-fold (P < 0.05) and 3.09-fold (P < 0.01), respectively. In addition, time to maximum concentration (T_max) was shortened by approximately 1.16 h (P < 0.01) and 0.58 h, respectively. The AUC from time zero to time t (AUC_0–t) of COA-I increased by 1.33-fold (P < 0.05) and 3.06-fold (P < 0.01) compared to PM-I and free DAP, respectively. In contrast, the bioavailability of PIO in the COA-I system did not significantly improve when compared to PM or PIO alone. The AUC_0–∞ for PIO showed a slight decrease compared to PM (P < 0.05), indicating that PIO did not benefit from increased bioavailability in the COA-I system.

3.3. Metabolomics analyses

Diabetes is characterized by abnormalities in the amino acid, fatty acid, glycerophospholipid, and sphingolipid metabolism. These early, broad-spectrum metabolic changes highlight the complexity of a disease predominantly defined by elevated blood GLU levels [24]. The early detection of such metabolic alterations is essential and should occur even before the manifestation of hyperglycemia manifests. Metabolomics was performed using normal diet mice and diabetic model mice. The normal diet mice were grouped into control, C-COA, and C-PM groups, while the diabetic model mice were assigned to MOD, M-COA, and M-PM groups. Each experiment was implemented in six groups in parallel (n = 6). Plasma metabolomics was performed following oral administration of samples from each group. Principal component analysis (PCA) (Fig. S2A) and orthogonal partial least squares-discriminant analysis (OPLS-DA) (Figs. S2B–G) were applied for untargeted metabolomic research. A pairwise comparison between the HFD diabetic group and the normal control group identified 2,949 potential different metabolites, of which 1,235 were upregulated and 1,714 were downregulated, involving 36 different metabolic pathways (Fig. S3). Moreover, a comparison between the HFD diabetic group and the normal control group revealed the dysregulated GLU and lipid metabolism pathways. For the normal diet mice, 2,975 potential different metabolites were identified between the C-PM and C-COA groups, with 2,077 upregulated and 898 downregulated metabolites, involving 27 different metabolic pathways (Fig. S4), suggesting that most of the GLU and lipid metabolism pathways were affected by the drug intervention. For the HFD diabetic models, 1,159 potential different metabolites were identified between the M-PM and M-COA groups, with 340 upregulated and 819 downregulated metabolites, involving 13 different metabolic pathways, suggesting that most of the disordered GLU and lipid molecules in the HFD diabetic group were reversed by the drug intervention (Fig. 3). The focus then shifted to comparing the M-PM and M-COA groups to identify changes in the synergistic mechanisms between COA-I and PM. The observed differences the in metabolites between the M-PM and M-COA groups constitute a fundamental basis for investigating potential pathways, regulatory mechanisms, and biomarkers. This approach may ultimately contribute to the advancement of more effective and personalized therapeutic strategies.

Fig. 3.

Metabolite profiles were analyzed in high-fat diet (HFD)-induced diabetic models, comparing the co-amorphous (M-COA) group to the physical mixture (M-PM) groups. (A) Volcano plot. (B) Kyoto Encyclopedia of Genes and Genomes (KEGG) classification. (C) KEGG enrichment. (D) Differential abundance score (DA score). (E) Pathway analysis. (F) Network analysis. (G) Schematic diagram of the metabolomics mechanism study of the sodium-dependent glucose transporters 2 (SGLT-2)/peroxisome proliferator-activated receptor-γ (PPAR-γ) dual receptor supramolecular system for the treatment of diabetes and obesity. AMPK: adenosine monophosphate-activated protein kinase; PI3K: phosphatidylinositol 3-kinase; AKT: protein kinase B; HIF: hypoxia inducible factor; Cer: ceramide; LPC: lysophosphatidylcholine; CerP: ceramide phosphate; S1P: sphinganine-1-phosphate; TRIG: triglyceride; PE: phosphatidylethanolamine; PS: phosphatidylserine; PC: phosphatidylcholine.

3.3.1. Visualization analysis of different metabolites

The Donut Plot revealed that substances involved in sugar and lipid metabolism accounted for 11.76% of fatty acids and 34.12% of lipids and lipid-like molecules, respectively, among the overall different metabolites between the PM and COA groups (Fig. S5A). The different metabolite screening is displayed in the matchstick plots (Fig. S5B), Kyoto Encyclopedia of Genes and Genomes (KEGG) heatmap (Fig. S5C), and hierarchical clustering analysis (Fig. S5D). The visualization analysis of the different metabolites between COA and PM indicated that COA induced different metabolomic features in diabetic mice when compared to PM (Fig. S5D). After performing a correlational analysis of different metabolites through chord plot analysis (Fig. S5E), it was found that the related pairs involved choline/CerP (d18:1/20:0), LPC (18:0/0:0)/TCA-ethyl, LPC (18:0/0:0)/TRIG (20:0/24:0/22:6(4Z,7Z,10Z,13Z,16Z,19Z)), 1-oleoyl-sn-glycerol 3-phosphate/glycerophosphocholine, and prostaglandin A1/cedronellone.

3.3.2. Metabolic pathway

The different metabolite screening is displayed in the heatmap of volcano plots (Fig. 3A). The KEGG classification chart (Fig. 3B) highlights the metabolic pathways impacted in the comparison between M-PM and M-COA. These include amino acid metabolism, lipid metabolism (e.g., sphingolipid, ether lipid, and glycerophospholipid metabolism), digestive system pathways, and the hypoxia inducible factor (HIF)-1 signaling pathway. Differential metabolites identified between the PM and COA-I groups were further analyzed using KEGG enrichment analysis (Fig. 3C).

The differential abundance score (DA Score) indicates the overall change in metabolites within a pathway. In the DA score plot, a score of −1 for glycerophospholipid metabolism, ascorbic acid, aldarate metabolism, glycine-serine-threonine metabolism, and the HIF-1-signaling pathway indicated downregulation of all annotated differential metabolites (Fig. 3D). Metabolic pathways analysis of differential metabolites was visualized using a bubble chart (Fig. 3E), revealing glycerophospholipid metabolism as a key metabolic pathway. The differences between PM and COA-I in the treatment of diabetes and obesity mainly originated from the differences in 13 metabolic pathways, with glycerophospholipid, ascorbic acid and aldarate, sphingolipid, and glycine-serine-threonine metabolism (related to AKT) pathways playing the key role. Network-based enrichment analysis was conducted using the differential metabolites identified between the PM and COA-I groups (Fig. 3F), resulting in metabolic pathways, modules, enzymes, reactions, and metabolites. In this study, the metabolomic mechanisms of the SGLT-2/PPAR-γ dual-receptor supramolecular system for the treatment of diabetes and obesity were explored from the following aspects (Fig. 3G).

3.3.2.1. Carbohydrate metabolism

Intervention with PM and COA-I resulted in reductions in the serum levels of ascorbic acid and GDP-4-amino-4,6-dideoxy-alpha-d-mannose, with COA-I displaying a more pronounced decrease when compared to PM. This downregulation explains the differences between PM and COA-I in carbohydrate metabolism, which further emphasizes COA-I’s stronger effect in modulating this pathway.

3.3.2.2. Amino acid metabolism

The main amino acid metabolisms involved in the PM and COA-I groups were those of glycine and serine metabolisms, and their differential metabolites choline increased in the diabetic model mice. Upon activation, AMPK decreased cholesterol synthesis and enhanced its efflux, thereby consequently reducing the risk of atherosclerosis [25]. After intervention with PM and COA-I, the choline content decreased in both treatment regimens, with a more significant reduction in the COA group than in the PM group. In the chord plot analysis of differential metabolite correlation, choline was correlated with ceramide phosphate (CerP) (d18:1/22:0).

3.3.2.3. Sphingolipid metabolism

Sphingolipid ceramides (Cer) can impair the activation and translocation of insulin-induced AKT kinase, thereby inducing insulin resistance [26]. Cer synthesis and increased levels enhance sterol regulatory element-binding proteins (SREBP)-mediated gene transcription and lipogenesis [27]. Cer can inhibit GLUT4 translocation to the cell membrane [28]. AKT and SREBP-1 proteins are the key downstream targets in the PI3K/AKT and AMPK pathways. When compared to the PM intervention, the levels of Cer (d20:1/20:3(6,8,11)-OH (5)) and CerP (d18:1/20:0) in the COA-I intervention group decreased significantly. The COA-I group exhibited decreased levels of TRIG (20:0/24:0/22:6 (4Z,7Z,10Z,13Z,16Z,19Z)), CerP (d18:1/22:0), and prostaglandin A1, suggesting its advantageous effects on lipid metabolism and inflammation modulation [29]. Selective reduction of sphinganine-1-phosphate (S1P) via the PI3K/AKT pathway enhances insulin sensitivity, mitigates myocardial hypertrophy, and prevents thrombosis [30]. Diabetic mice exhibited elevated S1P (d18:0), which was lowered by PM and further decreased by COA-I.

3.3.2.4. Glycerophospholipid metabolism

Lysophosphatidylcholine (LPC) and phosphatidylethanolamine (PE) (36:1) are associated with an increased risk of diabetes [31]. In the present study, the content of glycerophospholipid metabolism LPC (18:0/0:0), glycerophosphocholine, PE (18:0/20:4(6E,8Z,11Z,14Z) + = O (5)), and phosphatidylserine (PS 14:1(9Z)/18:2(10E,12Z) + = O (9)) were increased in the diabetic model mice but decreased after COA-I intervention.

The metabolic differences between PM and COA-I in HFD-induced diabetic mice were comprehensively analyzed, revealing significant variations in the activation of the PI3K/AKT and AMPK signaling pathways. Cermetabolism in the sphingolipid pathway emerged as a critical factor. To elucidate the mechanism of action of the DAP-PIO supramolecular system, future in vitro and in vivo studies will be conducted to evaluate and validate whether it reduces blood GLU levels and reverses hepatic steatosis and myocardial hypertrophy via the PI3K/AKT and AMPK signaling pathways.

3.4. In vitro cell pharmacological efficacy

3.4.1. Cytotoxicity

To investigate the mechanism of COA in mitigating insulin resistance in liver and heart tissues, HepG2 and H9c2 cell lines were employed. Supramolecular systems with varying component ratios were tested to assess their effects. The cytotoxicity of COA-I (Figs. 4A and B), COA-II (Figs. S6A and B), and COA-III (Figs. S6C and D) under HGHI conditions was evaluated using the MTT assay. As displayed, for both H9c2 and HepG2 cells, the COAs exhibited no significant cytotoxicity under HGHI conditions. However, PIO showed cytotoxicity to the cells at concentrations higher than 10 μM, while the survival rate of PM and the COAs slightly declined with increasing concentration. Therefore, a concentration less than 10 μM was selected for subsequent experiments.

Fig. 4.

H9c2 and HepG2 cells under hyperglycemic hyperinsulinemic (HGHI) conditions in the co-amorphous (COA)-I supramolecular system. The differential effects on H9c2 and HepG2 cells were studied across the following experimental groups: the control group under normal conditions, the group exposed to HGHI conditions (MOD), and the group of HGHI-induced cells treated with either DAP, PIO, a physical mixture (PM), or COA-I. (A, B) Cell viability rates of H9c2 (A) and HepG2 (B) cells under HGHI conditions in different groups. (C, D) Glucose (GLU) consumption rates of H9c2 (C) and HepG2 (D) cells under HGHI conditions in different groups. (E, F) Immunofluorescence of H9c2 (E) and HepG2 (F) cells GLU uptake under HGHI conditions in different groups. (G–J) Quantitative and statistical comparison of GLU uptake in H9c2 and HepG2 cells in different groups: GLU uptake in H9c2 cells (G), corresponding statistical analysis (H), GLU uptake in HepG2 cells (I), and corresponding statistical analysis (J). (K) Binding mode of dapagliflozin (DAP) with sodium-dependent glucose transporters 2 (SGLT-2) from the crystal structure. (L) The binding mode of the assembled molecule with SGLT2 from the docking results. (M) The superimposition of DAP and assembled molecule with SGLT2. (N) The binding mode of pioglitazone (PIO) with peroxisome proliferator-activated receptor-γ (PPAR-γ) from the docking results. (O) The binding mode of assembled molecule with PPAR-γ from the docking results. (P) The superimposition of PIO and assembled molecules with PPAR-γ. DAP and PIO are indicated with white sticks, and assembled molecules are indicated with yellow sticks. SGLT2 and PPAR-γ are indicated with green sticks and with cartoons, respectively. (2-deoxy-2-[(7-nitro-2,1,3-benzoxadiazol-4-yl) amino]-d-glucose (2-NBDG) (green), 4′,6-diamidino-2-phenylindole (DAPI) (blue), and merge in control.). Data are mean ± standard deviations (SD), n = 3. Significance levels were denoted as follows: ^∗P < 0.05, ^∗∗P < 0.01, ^∗∗∗P < 0.001, with these values indicating statistical significance.

3.4.2. GLU consumption rate and uptake assay

The effects of COAs on GLU metabolism in insulin-resistant HepG2 and H9c2 cells under HGHI conditions were assessed by using a GLU assay kit (Figs. 4C, 4D, and S6E–H). COA-I significantly enhanced GLU consumption when compared to PM, indicating improved cellular GLU utilization (Figs. 4C and D). This finding preliminarily confirmed the synergistic GLU-lowering effects of PIO and DAP in COAs. HepG2 cells exhibited greater sensitivity to the synergistic effects of PIO and DAP when compared to H9c2 cells (Fig. 4D). Similar results were observed for the COA-II systems (Figs. S6E and F). However, at concentrations exceeding 10 mM, GLU consumption decreased, likely due to the cytotoxic effects associated with higher API concentration. Therefore, selecting the appropriate concentrations of API is crucial for the optimal effect. The best effects were observed with the COA-I (DAP:PIO = 10 μM:10 μM), COA-II (DAP:PIO = 10 μM:5 μM), and COA-III (DAP:PIO = 5 μM:10 μM); therefore, these three ratios were selected for further studies on blood sugar reduction.

To elucidate the differences in GLU consumption among DAP, PIO, PMs, and COAs, qualitative and quantitative analyses of 2-NDBG uptake were performed on insulin-resistant HepG2 and H9c2 cells under HGHI conditions by using confocal microscopy and flow cytometry. As shown in Figs. 4E–J, GLU uptake by HepG2 and H9c2 cells was significantly higher under COA-I intervention compared to that under PM and pure API interventions, suggesting that COA-I effectively reduced cellular insulin resistance. Similar trends were observed for COA-II and COA-III systems (Fig. S7).

3.4.3. Synergistic mechanism of blood sugar reduction

Molecular docking analyses were performed to explore the synergistic mechanism of blood sugar reduction in supramolecular systems. Under conditions of HGHI, COAs demonstrated synergistic effects on GLU metabolism in insulin-resistant HepG2 and H9c2 cells. When small molecule exhibits biological activity, they typically initiate the process by specifically binding to their cognate receptors. A lower binding energy typically correlates with enhanced stability of this binding interaction, often serving as an indicator of superior biological potency. Considering that SGLT2 and PPAR-γ serve as the specific protein targets for DAP and PIO, respectively, molecular docking studies were performed using these proteins serving model systems to evaluate their interaction energies and stabilities. The present study aimed to elucidate the relationship between synergistic effects in supramolecular systems and intermolecular hydrogen bonding. The binding modes are depicted in Figs. 4K–P.

For the SGLT2 receptor protein, the binding mode of DAP is directly derived from the crystal structure, revealing a rich hydrogen-bond network formed between the hydroxyl groups on the DAP sugar unit and multiple amino acids within the active cavity of SGLT2, such as W291, S287, Q457, and H80, which stabilizes the binding of DAP (Fig. 4K). In the binding mode of the docked DAP-PIO assembly system with SGLT2, we observe that the two hydrogen bonds formed during the assembly of the two molecules remain stable. Moreover, the assembly forms hydrogen-bond interactions with Q457, D454, and F453, albeit not as extensive as the hydrogen-bond network formed between DAP and SGLT2. However, compared to the single DAP molecule, the PIO moiety of the assembly can penetrate deeper into the bottom of the active cavity, forming hydrophobic interactions with amino acids such as I297 and Y290 at the bottom (Figs. 4L and M). This allows for a better match with the shape of the active cavity. From the calculation results of the binding energy, we can also observe that the assembly binds more tightly to SGLT2 compared to DAP. The primary reason is that the van der Waals energy term of the assembly molecule (ΔG_vdw = −17.066 kcal/mol) is more favorable than that of DAP (ΔG_vdw = −14.753 kcal/mol), which is shown in Table S4.

For the PPAR-γ protein receptor, docking results show that the PIO molecule forms hydrogen-bond interactions with E259 and Q273 through the N-carbon atom and oxygen atom on its five-membered heterocyclic ring, respectively. Additionally, it forms hydrophobic interactions with amino acids such as L288 and L333, enabling its stable binding to PPAR-γ (Fig. 4N). However, when compared to the assembled molecule, the latter can penetrate deeper into the binding cavity of PPAR-γ. It forms hydrogen-bond interactions with Y473 within the cavity and π-π or T-π interactions with H449 and Y473, allowing it to occupy the active cavity more tightly and with a better fit (Figs. 4O and P). This is also evident from the binding energy calculations shown in Table S4, in which assembled molecule has better binding energy (ΔG_binding = −10.721 kcal/mol) with PPAR-γ than that of PIO (ΔG_binding = −7.270 kcal/mol). This may be the reason for the enhanced activity of DAP and PIO when they are combined through hydrogen bonding. At a mechanistic level, molecular docking has been employed to elucidate the relationship between intermolecular hydrogen bonds in supramolecular systems and their cellular efficacy. Previous research on gallic acid cocrystals has also demonstrated this point, namely, that the variations in α-glucosidase inhibitory activity between cocrystals and PMs can be attributed to these intermolecular hydrogen bonds [4]. These observations preliminarily suggest that the enhanced pharmacodynamic effects are attributable to the presence of inter-molecular hydrogen bonds.

The pathways implicated in the mechanism of blood sugar reduction: Among the three COAs, COA-I demonstrated superior GLU uptake and enhanced cellular function. Therefore, COA-I was selected to investigate the mechanism pathways underlying blood sugar reduction. The metabolomics analysis results presented in Section 3.3 indicate that COA-I has the potential to enhance the activity of related biomarkers, regulate metabolic processes, and reduce blood GLU and lipid levels through the AMPK and PI3K/AKT signaling pathways. In terms of metabolic differences, a significant downregulation was observed in Cer, a key metabolite in sphingolipid metabolism. Cer inhibits the translocation of GLUT4 to the cell membrane [32]; thus, its decreased level effectively enhances GLUT4-mediated GLU uptake capacity. In order to further verify this result, the effect of COA-I on the expression of GLUT4 in HepG2 and H9c2 cells with insulin resistance was investigated. As shown in Fig. 5, the GLUT4 expression was significantly downregulated in insulin-resistant cells but was restored following COA-I intervention. To explore the involvement of signaling pathways, PI3K/AKT and AMPK pathway inhibitors, LY294002 and compound C, respectively, were used. As depicted in Figs. 5A–D, both the inhibitors attenuated the strong GLUT4 fluorescence induced by COA-I, indicating that COA-I’s effect was mediated through the activation of the PI3K and AMPK pathways. These findings suggest that COA-I alleviated insulin resistance in HepG2 and H9c2 cells by enhancing the PI3K/AKT and AMPK pathway activities (Fig. 5E).

Fig. 5.

The pathways implicated in the mechanism of blood sugar reduction involved H9c2 and HepG2 cells under hyperglycemic hyperinsulinemic (HGHI) conditions when treated with dapagliflozin-pioglitazone (DAP-PIO) co-amorphous (COA)-I. The differential effects on H9c2 and HepG2 cells were studied across the following experimental groups: the control group under normal conditions, the group exposed to HGHI conditions (MOD), and the group of HGHI-induced cells treated with either DAP, PIO, a physical mixture (PM), or COA-I. Inhibitor LY294002 and dorsomorphin (compound C) were pretreated for the examination of the underlying pathways involved in the mechanism. (A) Representative immunofluorescence images showing glucose transporter 4 (GLUT4) (green) and nuclei stained with 4′,6-diamidino-2-phenylindole (DAPI) (blue) in H9c2 cells cultured under high-glucose/high-insulin (HGHI) conditions across different treatment groups. (B) Representative immunofluorescence images of GLUT4 (green) and DAPI (blue) in HepG2 cells under HGHI conditions across different groups. (C) Quantitative analysis of GLUT4 expression levels in H9c2 cells. (D) Quantitative analysis of GLUT4 expression levels in HepG2 cells. (E) The process of GLUT4 expression through the phosphatidylinositol 3-kinase (PI3K)/protein kinase B (AKT) and adenosine monophosphate-activated protein kinase (AMPK) pathways is induced by the supramolecular system. Data are mean ± standard deviations (SD), n = 3. Significance levels were denoted as follows: ^∗P < 0.05, ^∗∗P < 0.01, and ^∗∗∗P < 0.001, with these values indicating statistical significance. IR: insulin receptor; IRS: insulin receptor substrate.

3.5. Pharmacological efficacy evaluation in vivo

3.5.1. Hypoglycemic effect and improvement on insulin sensitivity

It is precisely due to the interaction between PIO and DAP molecules that COA-I not only improves the solubility, dissolution rate, and bioavailability of the APIs, but also enhances its GLU consumption and uptake capacity. The hypoglycemic effects of COA-I in vivo were therefore attributed to these enhancements. After 3 months of HFD feeding, the FBG levels in HFD-fed mice were significantly elevated (≥11.1 mmol/L), which confirmed the successful establishment of the T2DM mouse model (Fig. 6A). The hypoglycemic effects of COA-I were evaluated using OGTT and ITT. In the HFD mouse model, after treatment with different preparations, the OGTT results displayed the blood GLU change curve over time (Fig. 6B) and its corresponding AUC (Fig. 6C), while the ITT results showed the blood GLU change curve (Fig. 6D) and its AUC (Fig. 6E). When compared to the control group, diabetic mice in the model (MOD) group exhibited consistently elevated blood GLU levels. Treatment with PIO, DAP, PM, and COA-I significantly reduced blood GLU levels in diabetic mice (Figs. 6B and D). Moreover, COA-I demonstrated superior global GLU tolerance in OGTT and ITT assays when compared to PM, as reflected by reduced AUC values (P < 0.05) (Figs. 6C and E).

Fig. 6.

The normal diet (ND)-wild-type (WT) group, high-fat diet (HFD)-WT diabetic mice group (MOD), and HFD-induced diabetic mice treated with either dapagliflozin (DAP), pioglitazone (PIO), a physical mixture (PM), or co-amorphous (COA-I) exhibited varying effects on glucose (GLU) homeostasis and insulin resistance. (A) The scheme for the establishment and treatment of the HFD-induced diabetic mice model. (B) The blood GLU curve of oral glucose tolerance test (OGTT). (n = 6). (C) The area under the curve (AUC) for GLU in OGTT (n = 6). (D) The blood GLU curve of insulin tolerance test (ITT) (n = 6). (E) AUC for GLU in ITT (n = 6). (F) The serum triglyceride (TRIG) levels (n = 3). (G) The total cholesterol (TC) levels (n = 3). (H) The high-density lipoprotein cholesterol (HDL-C) levels (n = 3). (I) The low-density lipoprotein cholesterol (LDL-C) levels (n = 3). (J) Serum quantification of alanine aminotransferase (ALT) (n = 3). (K) Serum quantification of aspartate aminotransferase (AST) (n = 3). (L) The serum fasting insulin levels (FINS) (n = 3). (M) Homeostasis model assessment-insulin resistance (HOMA-IR) index (n = 3). (N) Hematoxylin and eosin (H&E) of the pancreatic sections. (O) Immunohistochemistry staining for insulin (shown in tan). (P) Immunofluorescence staining of insulin (shown in green fluorescence). Data are mean ± standard deviations (SD). Significance levels were denoted as follows: ^∗P < 0.05, ^∗∗P < 0.01, and ^∗∗∗P < 0.001, with these values indicating statistical significance.

Furthermore, serum biochemical markers were also measured (Figs. 6F–K). Lipid profiles in COA-I-treated mice displayed decreased TC, TRIG, and LDL-C levels when compared to PM-treated mice, alongside an increase in the HDL-C level compared to that in the MOD group. Statistically significant differences were observed between the PM and COA-I groups in terms of the TC, TRIG, and LDL-C levels. The integration of biochemical indices with metabolomics data demonstrated a complementary relationship. For instance, our study revealed that HFD caused a significant elevation in the blood lipid levels in mice; these findings aligned with increases in metabolites such as TRIGs and cholesterol in the metabolomics data. This complementarity not only enhances our comprehension of the metabolic alterations induced by an HFD but also provides valuable insights and avenues for further investigation into the mechanisms underlying metabolic regulation. Liver function markers, ALT and AST, were measured to evaluate the effects of treatments on liver health (Figs. 6J and K). The ALT and AST levels were elevated in the MOD group compared to that in the control group, indicating liver dysfunction in T2DM mice. Treatment with different interventions reduced the ALT and AST levels to varying degrees, with the COA groups showing reductions comparable to those in the control group.

Insulin sensitivity was assessed in HFD-fed mice using the HOMA-IR index (Figs. 6L and M). The plasma insulin levels were significantly higher in the MOD group than in the control group, reflecting pronounced insulin resistance. Treatment with the various interventions of different groups of samples reversed the high insulin levels and high HOMA-IR index, especially in the COA-I group. More importantly, the COA-I group showed significantly lower values of insulin and HOMA-IR index compared to the PM group (P < 0.05). COA-I treatment effectively lowered FBG and insulin levels and improved GLU tolerance and peripheral insulin resistance, as demonstrated by OGTT, ITT, and HOMA-IR results. Insulin resistance often impairs pancreatic insulin secretion due to β-cell dysfunction. To evaluate tissue-level changes, major organs (such as the heart, liver, pancreas, spleen, and kidney) from different treatment groups were subjected to H&E staining, with specific analyses of the pancreas, liver, and heart. As depicted in Fig. S8, staining results of spleen and kidney tissues revealed intact tissue structure with no apparent abnormalities. Pancreatic H&E staining (Fig. 6N) revealed intact, regularly distributed islets with abundant pancreatic β-cells and normal acinar cells in the control group. In contrast, islets in MOD diabetic mice appeared atrophic and severely damaged, with significant reductions noted in the in β-cell numbers, infiltration of acinar cells into the islets, vacuolation, and abnormal acinar structures. After 3 weeks of treatment, pancreatic islet pathology and β-cell counts improved significantly across all intervention groups, with the COA group demonstrating superior recovery when compared to the PM group. Acinar cell morphology also showed marked improvement in the COA-treated groups. Similarly, immunohistochemistry (Fig. 6O) and fluorescent staining (Fig. 6P) of pancreatic tissues supported these findings. In conclusion, COA-I interventions demonstrated significant improvements in both exocrine and endocrine pancreatic functions, which effectively enhanced insulin sensitivity.

3.5.2. The mechanism of reverse hepatic steatosis and insulin resistance

HFD-induced hepatocellular swelling, cellular disarray, lipid accumulation, and both micro vesicular and macro vesicular lipid deposition in the liver, which increased the liver volume [24]. AKT activation is known to inhibit hepatic GLU production, which contributes to hypoglycemic effects [33]. In the COA-I treatment group, increases in liver weight (Fig. 7A) and the liver index (Fig. 7B) were effectively prevented. To investigate the mechanisms underlying COA-I’s hypoglycemic and lipid-lowering effects, the PI3K/AKT and AMPK pathways—which are the key regulators of GLU and lipid metabolism—were examined in the livers of HFD-fed mice. The mRNA levels of PI3K, AKT, and AMPK were significantly upregulated in the COA-I group compared to that in the MOD and PM groups (Figs. 7C–E; primer sequences are detailed in Table S1). The activation of the PI3K/AKT and AMPK pathways, which are important regulators of cellular energy, played a significant role in GLU metabolism regulation. The results from Western blot analysis in this study agreed with the mRNA expression trends of PI3K/AKT and AMPK, as determined by qRT-PCR (Figs. 7F–I). Compared to the control group, the levels of p-AKT, p-PI3K, and p-AMPK were reduced in the liver of the HFD group. In contrast, interventions with different groups' samples restored their expression, suggesting that DAP and PIO may prevent liver fat degeneration-related insulin resistance and the reduction of hepatic glycogen synthesis. Furthermore, the advantages of reverse hepatic steatosis and insulin resistance were noted in the COA-I group relative to that in the PM group, highlighting the benefits of improved bioavailability of COA-I.

Fig. 7.

The effects of the dapagliflozin-pioglitazone (DAP-PIO) co-amorphous (COA)-I on the liver tissues in high-fat diet (HFD) mice. (A) The differential impact on liver weight was studied across the following groups: the normal diet (ND)-wild-type (WT) group (control group), HFD-WT diabetic mice group (MOD), and HFD-induced diabetic mice treated with either PIO, DAP, a physical mixture (PM), or COA-I (n = 6). (B) Liver index (expressed as liver weight/body weight, n = 6). (C–E) The messenger RNA (mRNA) expression of phosphatidylinositol 3-kinase (PI3K) (C), protein kinase B (AKT) (D), and adenosine monophosphate-activated protein kinase (AMPK) (E) in the mouse liver (n = 3). (F–I) Protein levels in mouse liver: the immunoblot images (F) and quantitative analysis show the expression levels of phosphorylated PI3K (p-PI3K) (G), phosphorylated AKT (p-AKT) (H), and phosphorylated AMPK (p-AMPK) (I) in mouse liver tissues (n = 3). (J) Hematoxylin and eosin (H&E) staining. (K) Oil red O staining. (L) Periodic acid-Schiff (PAS) staining of mouse liver sections in each group. (M) The process of hepatic glycogen through the PI3K/AKT pathway is induced by the supramolecular system. Data are presented as mean ± standard deviations (SD). Significance levels were denoted as follows: ^∗P < 0.05, ^∗∗P < 0.01, and ^∗∗∗P < 0.001, with these values indicating statistical significance. IR: insulin receptor; IRS: insulin receptor substrate; GLUT4: glucose transporter 4; T2DM: type 2 diabetes mellitus; STZ: streptozocin; GSK: glycogen synthase kinase.

The effects of the DAP-PIO COA on liver histopathology were further examined using H&E staining (Fig. 7J). The control group mice had normal liver tissues, while the model group mice exhibited damaged liver structure with misaligned hepatocytes, unclear intercellular spaces, and lipid accumulation. Post-intervention, the liver structure in the PM and COA-I groups improved significantly, particularly in the COA group. Liver Oil red O staining further supported these findings (Fig. 7K), with the corresponding semi-quantitative data are presented in Fig. S9A. PAS staining revealed that COA-I could improve the reduction of glycogen synthesis induced by prolonged HFD feeding (Figs. 7L and M), indicating a protective effect on the liver tissue damage in diabetic mice. Compared to the PM, the COA-I group showed advantages in regulating the PI3K/AKT and AMPK pathways in HFD-induced diabetic mice, which play a significant role in GLU and lipid metabolism regulation. In COA-I, DAP and PIO synergistically acted on the PI3K/AKT and AMPK pathways.

3.5.3. The mechanism of ameliorates myocardial hypertrophy and insulin resistance amelioration

HFD-induced sustained for approximately 3 months can induce diabetic cardiomyopathy and myocardial insulin resistance [34], as characterized by myocardial hypertrophy, which leads to increases in the heart index, cardiomyocyte size, and left ventricular mass [35]. Treatment with COA-I mitigated these effects, as evidenced by reductions in the heart weight and heart index in HFD-fed mice (Figs. 8A and B). To elucidate the regulatory role of COA-I on GLU and lipid metabolism, the PI3K/AKT and AMPK signaling pathways were analyzed in the hearts of HFD-fed mice. qRT-PCR revealed suppressed mRNA levels of PI3K, AKT, and AMPK in the model group. However, COA-I treatment significantly increased their expression, surpassing other treatment groups and approaching levels seen in the control group (Figs. 8C–E). Western blotting confirmed these findings, indicating that COA-I elevated the expression of phosphorylated proteins p-PI3K, p-AKT, and p-AMPK to levels comparable to those of the control group, demonstrating superiority over PM treatment (Figs. 8F–I).

Fig. 8.

The effects of dapagliflozin-pioglitazone (DAP-PIO) co-amorphous (COA)-I on the cardiac tissues in high-fat diet (HFD) mice. (A) The study assessed the differential impact on heart weight was studied across the following groups: the normal diet (ND)-wild-type (WT) group (control group), HFD-WT diabetic mice group (MOD), and HFD-induced diabetic mice treated with either PIO, DAP, PIO, a physical mixture (PM), or COA-I. n = 6. (B) Cardiac index (expressed as heart weight/body weight, n = 6). (C–E) The messenger RNA (mRNA) expression of phosphatidylinositol 3-kinase (PI3K) (C), protein kinase B (AKT) (D), and adenosine monophosphate-activated protein kinase (AMPK) (E) in the mouse heart. n = 3. (F–I) Protein levels in mouse heart: the immunoblot images (F) and quantitative analysis show the expression levels of phosphorylated PI3K (p-PI3K) (G), phosphorylated AKT (p-AKT) (H), and phosphorylated AMPK (p-AMPK) (I) in mouse heart tissues (n = 3). (J) Hematoxylin and eosin (H&E) staining. (K) Oil red O staining. Data are presented as mean ± standard deviations (SD). Significance levels were denoted as follows: ^∗P < 0.05, ^∗∗P < 0.01, and ^∗∗∗P < 0.001, with these values indicating statistical significance.

H&E staining (Fig. 8J) revealed elongated, regularly arranged, and tightly packed cardiomyocytes in the control group. In contrast, HFD-induced cardiomyopathy in the model group resulted in significant edema, hypertrophy, vacuolar degeneration. Additionally, Oil red O staining showed increased lipid content in the cardiac tissues of HFD mice (Fig. 8K), with the corresponding semi-quantitative data presented in Fig. S9B. In comparison to the PM group, the COA-I group demonstrated superior benefits in enhancing cardiac tissue structure, as well as in mitigating lipid degeneration and inflammation. In summary, COA-I alleviated myocardial insulin resistance and myocardial hypertrophy in HFD-induced diabetic mice by activating the PI3K/AKT and AMPK signaling pathways.

4. Discussion

Pharmaceutical supramolecular APIs, including the COC and COA forms, can modify the physicochemical properties [36]. In this study, the DAP-PIO supramolecular system was successfully screened and characterized. The results demonstrated that this supramolecular system could significantly enhance the hypoglycemic effect in vivo following therapeutic application. This improvement can be attributed to two key factors: first, the optimization of bioavailability, and second, the synergistic interaction between the two components within the supramolecular system. Specifically, bioavailability studies demonstrated that the DAP component within the supramolecular system exhibited enhanced bioavailability, while the bioavailability of the PIO component did not increase. Notably, not all supramolecular systems can increase bioavailability for both components; in many cases, only one component’s bioavailability is enhanced, and a reduction may be seen in the other. For example, it has been documented that the COC formed by piroxicam and 4-aminobenzoic acid enhances the bioavailability of piroxicam, but decreases that of 4-aminobenzoic acid [37]. Similarly, the COC formed by pyrazinamide (PZA) and baicalin (BCL) does not lead to an increase in the C_max of PZA components. In such supramolecular systems, the differential effects of hydrogen bonding on each component can decrease the bioavailability of one component while enhancing that of another [38]. In the case of PIO, despite the decrease in its bioavailability within COA-I, there was no negative impact on the overall hypoglycemic effect in vivo. The potential reasons for the aforementioned observation stem from PIO’s robust activity, its unique action mechanism, and its stable metabolism [39]. Among, these metabolites continue to possess the ability to activate PPAR-γ and other key signal transduction pathways, potentially sustaining their beneficial impacts on physiological functions linked to the drug’s effectiveness, such as blood GLU regulation, despite the reduced bioavailability of the parent compound. Consequently, the fluctuating bioavailability of PIO, fueled by the presence and persistent activity of these metabolites, does not invariably lead to corresponding changes in its efficacy [39].

Pharmaceutical supramolecular of APIs can induce the synergistic efficacy of two components, however, the in vivo mechanism remains unclear. Direct analyses of the supramolecular weak interaction between the two components and its effects in the presence of a multitude of complex biological substances are rather difficult. As such, the reports on the synergistic mechanism of the two components in the supramolecular system are rare. Here, a metabolomics-driven approach was used in conjunction with biological validation, we aimed to elucidate the mechanisms underlying the synergistic action of two components in a supramolecular system. Subsequently, metabolomics technology was applied to analyze the differential metabolites arising from the absorption of DAP/PIO simple PM versus DAP-PIO supramolecular into the blood. These metabolite differences are deemed vital for understanding the treatment synergy, which can then help guide further research into pathways, regulatory factors, and biomarkers.

Carbohydrate metabolism: In this study, MOD mice exhibited increased serum levels of ascorbic acid and GDP-4-amino-4,6-dideoxy-alpha-d-mannose, which decreased after DAP and PIO treatment, with a greater reduction noted in the COA-I treatment group. Ascorbate aids collagen synthesis, which affects HIF-1α [40], influences fibrosis, and is associated with GDP-4-amino-4,6-dideoxy-alpha-d-mannose metabolism [41]. This association elucidates the observed variations in carbohydrate metabolism between PM and COA-I.

Amino acid metabolism: Amino acid metabolism is crucial in diabetes, particularly involving glycine, serine, and choline. Glycine protects diabetic β-cells from oxidative stress by boosting glutathione (GSH) synthesis through glycine transporter-1 and by reducing ROS via glycine receptors. AMPK, which is activated during energy deficiency, lowers cholesterol synthesis and increases cholesterol efflux and transport, thereby reducing intracellular cholesterol and atherosclerosis risk [25]. After intervention with PM and COA-I, the choline content decreased in both treatments, with a more significant reduction noted in the COA-I group than in the PM group. These metabolites were deemed vital in the PM and COA-I-intervention groups.

Sphingolipid metabolism: Sphingomyelin contributes to obesity, insulin resistance, and tissue growth. Ceramide, a type of sphingolipid, disrupts the metabolism by impairing mitochondria and promoting inflammation, resulting in insulin resistance [42]. Managing ceramides can alleviate insulin resistance by targeting AKT and SREBP-1 proteins [43]. AKT is a central regulator of insulin signaling that promotes GLU uptake and metabolism, while SREBP-1 is involved in lipid synthesis and homeostasis. By targeting these proteins, interventions can potentially restore insulin sensitivity and improve overall metabolic status [44]. The differences between PM and COA-I in sphingolipid metabolism revealed the superiority of COA-I to PM in activating the PI3K/AKT and AMPK pathways and in improving insulin resistance and lipid metabolic disorder. In the presence of excessive lipids, sphingolipids behave as lipotoxic lipids [45]. Sphingosine, S1P, and lysosphingomyelin increase PS synthesis, which potentially causes thrombosis. The selective reduction of S1P via the PI3K/AKT pathway enhances insulin sensitivity, mitigates myocardial hypertrophy, and prevents thrombosis [30]. Diabetic mice exhibited elevated S1P (d18:0), which was lowered by PM and further reduced by COA-I.

Glycerophospholipid metabolism: LPC triggers inflammation and atherosclerosis and is associated with diabetes [46]. LPC variants induce Interleukin-1 beta (IL-1β) release [47]. In the present study, the content of glycerophospholipid metabolism LPC (18:0/0:0), glycerophosphocholine, PE (18:0/20:4(6E,8Z,11Z,14Z) + = O (5)) and PS (14:1(9Z)/18:2(10E,12Z) + = O (9)) increased in diabetic model mice, but decreased after COA-I intervention.

COA-I was found to improve biomarkers, regulate metabolism, and lower blood sugar and lipid levels through the AMPK and PI3K/AKT pathways. GLUT4 enhanced the GLU uptake, with Cer acting as the key metabolite. In order to further verify this result, the effect of COA-I on the expression of GLUT4 in HepG2 and H9c2 cells with insulin resistance was explored. The study revealed that, under HGHI conditions, AMPK was inactive in HepG2 and H9c2 cells, but was activated by COA-I, which boosted the GLU metabolism through the insulin-signaling pathway. The AMPK inhibitor compound C confirmed the AMPK’s involvement. COA-I increased GLU consumption and GLUT4 expression; these effects were blocked by the PI3K inhibitor LY294002, indicating that COA-I promoted GLU use via a PI3K-dependent mechanism. COA-I improved the HGHI-impaired GLUT4 expression, which was reduced by LY294002 and compound C. The differences between PM and supramolecular forms were observed in activating the PI3K/AKT and AMPK pathways and in reducing hepatic steatosis and cardiac hypertrophy.

Mice fed with HFD over a long-time developed insulin resistance in peripheral and visceral tissues, as characterized by hepatic steatosis and increased left ventricular mass, respectively [34]. In the present study, a diabetic mouse model was established after a 3-month HFD feeding-induced procedure. In these mice, treatment with COA-I could significantly reduce the FBG and insulin levels. The results of OGTT, ITT, and HOMA-insulin resistance indicated that COA-I relieved GLU tolerance and peripheral insulin resistance in HFD mice. Histological and pathological analyses revealed that COA-I mitigated liver volume and left ventricular mass increases in HFD mice along with the preservation of the liver glycogen and reduced lipid accumulation in the HFD mice. Insulin resistance, which is linked to impaired glycogen synthesis [48], can be countered by AKT activation, as it lowers GLU production [49]. The PI3K/AKT pathway is thus deemed crucial for insulin function and GLU metabolism, and impaired AKT phosphorylation suggests GLU metabolism dysfunction and insulin-resistance severity [50]. In insulin-resistant diabetic mice, COA-I restored glycogen storage by activating the PI3K/Akt/glycogen synthase kinase (GSK) 3β pathway, as confirmed by the results of PAS staining.

In conclusion, the supramolecular system played a synergistic role in activating the PI3K/AKT and AMPK pathways to improve insulin sensitivity and regulate lipid metabolism, which facilitated relieving of hepatic steatosis and cardiac hypertrophy. The results provided an innovative strategy to develop the SGLT-2/PPAR-γ-dual receptor supramolecular system for enhancing the therapeutic efficacy against diabetes and obesity.

5. Conclusion

A novel SGLT-2/PPAR-γ dual receptor supramolecular system displaying synergistic therapeutic effects on diabetes and obesity was developed. The metabolomics approach was used to explore the synergistic mechanism of the components in the supramolecular system. DAP-PIO supramolecular system demonstrated significant advantages in improving the adverse physicochemical properties of the phlorizin analog DAP. The differential metabolites and metabolic pathways between PM and supramolecular in HFD-induced diabetic mice were comprehensively analyzed to reveal that the supramolecular system played a synergistic role in activating the PI3K/AKT and AMPK pathways toward improving insulin sensitivity and regulating lipid metabolism, resulting in relieving hepatic steatosis and cardiac hypertrophy. Our findings not only provide a holistic understanding of the impacts of the SGLT-2/PPAR-γ dual receptor supramolecular system on diabetes and obesity but also emphasize the significance of integrative metabolomics and biological technology in elucidating the synergistic therapeutic mechanisms of the two components in the supramolecular system.

CRediT authorship contribution statement

Saisai Ren: Writing – original draft, Visualization, Supervision, Resources, Methodology, Investigation, Formal analysis, Conceptualization. Han Hao: Formal analysis, Data curation, Conceptualization. Wei Guo: Software, Methodology. Mo Zhang: Supervision, Methodology, Funding acquisition. Honglin Feng: Methodology. Jing Wang: Writing – review & editing, Resources.

Declaration of competing interest

The authors declare that there are no conflicts of interest.

Appendix A. Supplementary data

The following are the supplementary data to this article: