All experimental procedures involving db/db and db/m mice were conducted in accordance with the ARRIVE guidelines and were approved by the Institutional Animal Care and Use Committee of Hubei University of Medicine (Shiyan, China; (approval no. 2024S071). Humane endpoints were as follows: i) Body weight loss ≥20% within 48 h (relative to peak body weight); ii) severe hypoglycemia (<40 mg/dl) or hyperglycemia (>600 mg/dl) refractory to clinical stabilization (for example, dietary adjustment or insulin administration); iii) functional impairment (inability to access food/water independently or lethargy persisting >24 h) and iv) overt signs of distress (labored breathing, hunched posture lasting >12 h or ulcerative lesions affecting >5% body surface area). No animals met criteria for early euthanasia. Animals were humanely euthanized via intraperitoneal injection of sodium pentobarbital at a dose of 150 mg/kg body weight. Animal death was confirmed by the absence of heartbeat, breathing and corneal reflex. These methods align with the American Veterinary Medical Association guidelines to ensure minimal animal suffering (13).

Key genes associated with DCM were investigated using The National Center for Biotechnology Information Gene Expression Omnibus (GEO) database (ncbi.nlm.nih.gov/geo/). A total of three rat transcriptome expression profiling datasets (accession nos. GSE4745, GSE5606 and GSE6880) were retrieved from the GEO database (Table I). To eliminate batch effects and other sources of variance, surrogate variable analysis (14) and limma packages (version 3.54.2, http://bioconductor.org/packages/release/bioc/html/limma.html) were employed in R (version 4.2.3; http://cran.r-project.org/bin/windows/base/old/4.2.3/) to merge the GSE4745 and GSE6880 datasets, and the resulting merged dataset served as the internal training set. For independent validation, GSE5606 dataset as an external validation cohort. This strategic approach of combining GSE4745 and GSE6880 not only increased statistical power but also maintained platform diversity within the training set. Importantly, all three datasets investigated identical species and disease models, ensuring biological consistency while allowing rigorous assessment of model generalizability across different experimental platforms.

GEO datasets were corrected using the limma package. The expression data files of the combined dataset were investigated for deviation analysis on the basis of |log2 (fold-change)|≥1 and adjusted P-value <0.05. DEGs were visualized using heatmaps generated by the heatmap package (version 1.0.12; cran.r-project.org/web/packages/pheatmap/index.html) and volcano plots created with the ggplot2 package (version 3.5.1; http://ggplot2.tidyverse.org/).

Gene Ontology (GO) (15) and Kyoto Encyclopedia of Genes and Genomes (KEGG) (16) pathway enrichment analyses were performed using the clusterProfiler R package (version 4.7.1; bioconductor.org/packages/release/bioc/html/clusterProfiler.html). GO enrichment analysis provides a comprehensive description of the gene and protein functions associated with biological processes (BPs), molecular functions and cellular components. Adjusted P<0.05 was considered to indicate a statistically significant difference. For the Gene Set Enrichment Analysis (GSEA), the clusterProfiler package (version 4.7.1; bioconductor.org/packages/release/bioc/html/clusterProfiler.html) was utilized with adjusted P<0.05 to indicate statistically significant enrichment. The results of enrichment analyses were visualized using the ggplot2 package (version 3.5.1; ggplot2.tidyverse.org/), and chord plots were generated using the Xian Tao Academic Online Analysis platform (xiantaozi.com/products/apply/c0b6febb-52dd-4525-970a-61bbe9e263ff/analyse/f9496965-0e1a-4249-b123-e6a00c924021).

WGCNA is a bioinformatics tool that calculates gene expression levels and clusters genes with similar expression patterns into modules. These functional gene clusters are used to explore interactions between genes and biological phenotypes within the modules (17). The present study employed the WGCNA package (version 1.72; cran.r-project.org/web/packages/WGCNA/index.html) to identify gene modules with similar expression patterns by constructing a gene co-expression network using the mean chained hierarchical clustering algorithm. These modules were further distinguished through hierarchical clustering based on proximity. Additionally, the association of each module was inferred by calculating the association between different modules and biological phenotype. Finally, core genes from key modules were identified. Diagnostic potential was evaluated by ROC curve analysis (AUC >0.7), with differential expression visualized through boxplots (Wilcoxon test, P<0.05).

Expression of DEGs intersecting with modal genes significantly associated with DCM, as identified through WGCNA, were analyzed. Intersection analysis was performed using the Venn package (version 1.12; cran.r-project.org/web/packages/venn/index.html) The GLMNET package (version 4.1; cran.r-project.org/web/packages/glmnet/index.html) was employed to construct a LASSO regression model for the intersected genes and DCM feature genes were identified through cross-validation. The genes corresponding to the points with the lowest cross-validation error were designated key genes associated with DCM.

To analyze the expression of key genes in both internal and external datasets, the Limma package was used. P<0.05 was considered to indicate a statistically significant difference. Additionally, the ggpubr package (version 0.6.0; http://cran.r-project.org/web/packages/ggpubr/index.html) was employed to create a diverging box plot to visualize significant differences in key gene expression levels between the disease and control groups.

ROC curves were used to validate the accuracy of key genes in diagnosing DCM. R software (version 4.2.3; cran.r-project.org/bin/windows/base/old/4.2.3/) was used to analyze key gene expression and the pROC (version 1.18.5; http://xrobin.github.io/pROC/) package was used to calculate and plot the area under the curve (AUC), with a larger AUC indicating a greater diagnostic value of the key gene.

The HL-1 murine cardiomyocyte cell line (Unico) was maintained at 37°C in a humidified atmosphere of 5% CO₂, 21% O₂ and 74% N₂. The cells were maintained in high-glucose DMEM (cat. no. SH30022.01; Cytiva) supplemented with 10% fetal bovine serum (FBS; Cat. No. SA101.02, Cellmax, Beijing, China). The medium was changed every 2 days, and cells in mid-logarithmic growth phase were collected for experiments. Cultured HL-1 cells were divided into low- (5.5 mM) and high-glucose (33.3 mM) groups, and incubated at 37°C for 48 h (18,19).

All experimental protocols were approved by the Animal Ethics Committee of Hubei University of Medicine (Shiyan, Hubei, China). A total of six male db/db diabetic mice (age, 6 weeks; body weight 27.5±1.6 g at study initiation; Certificate No. 202341294-202341295, Jiangsu Huachuang Sino Pharmaceutical Technology Co., Ltd, China) were used to establish a model of DCM, with six db/m mice serving as controls. Mice were maintained under a 12/12-h light/dark cycle at a controlled temperature of 22±1°C and 50±10% relative humidity, with ad libitum access to food and water. All animals were acclimatized for 2 weeks prior to experiments. Body weight and blood glucose levels were measured. Blood glucose levels were monitored via tail vein sampling. Briefly, the distal 1–2 mm of the tail tip was sterilized with 70% ethanol and air-dried. A sterile lancet was used to make an incision, and a single blood droplet (~1 µl) was collected. Glucose concentration was analyzed using a Sinocare Blood Glucose Monitoring System (Sinocare Inc., Changsha). Triplicate measurements were averaged per mouse to ensure reliability. All measurements were performed between 9:00 and 11:00 AM under non-fasting conditions to minimize circadian variation. In longitudinal studies, sequential sampling was restricted to the same tail region to avoid tissue damage. The animals were monitored in SPF-grade animal laboratories for 24 weeks, with echocardiograms performed every 4 weeks. Sampling was conducted at 32 weeks of age.

HL-1 cardiomyocytes were transduced with recombinant lentiviral particles carrying the human HK2 gene (Lenti-HK2; Vigene Biosciences, cat. no. FC-6073) following the manufacturer's recommended protocol. HL-1 cells were seeded in 12-well plates at 1×10⁵ cells/well and cultured at 37°C until reaching 40–50% confluence. Cells were maintained at 37°C in medium, which was replaced with 420 µl fresh complete DMEM supplemented with 10% FBS [cat. no. SA101.02; CellTech (Beijing) Co., Ltd.], supplemented with 20 µl LV-Enhance (50X, Vigene Biosciences) and 60 µl/ml lentivirus overexpressing human HK2 (LV-hHK2-OE) (Vigene Biosciences, cat. no. FBE0001, Shandong, China; http://www.wzbio.com.cn/). LV-ZsGreen-Puro-CON (Vigene Biosciences) served as negative controls. Subsequent experiments were performed 48 h post-transfection. The transfection was conducted at 37°C in a 5% CO₂ atmosphere for 18–24 h, followed by the addition of 500 µl complete DMEM (supplemented with 10% FBS [cat. no. SA101.02; CellTech (Beijing) Co., Ltd.]) for an additional 24 h under the same culture conditions. Green fluorescence was monitored for 24–48 h post-transfection to assess transfection efficiency. Once green fluorescence was detected in the cells, puromycin (3 g/ml; cat. no. HY-K1057; MedChemExpress) was added to each well, and the same concentration of puromycin was maintained for 7 days at 37°C with 5% CO. The cells were collected for further experiments and cryopreserved (−196°C) in liquid nitrogen.

Total protein was extracted from heart tissue and HL-1 cells using RIPA lysis buffer (Servicebio Technology Co., Wuhan, China; cat. no. G2002) containing 50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 1 mM EDTA-2Na, 1% sodium deoxycholate, 0.1% SDS and 1% Triton X-100. Protein concentration was determined using the BCAk method. Equal amounts of protein (20 µg/lane) were separated by 10% SDS-PAGE and transferred to a PVDF membrane (Immobilon-PSQ Blotting Membranes; 0.2 µm; Merck KGaA). Following 1-hour blocking at room temperature with 5% skimmed milk, the membranes were incubated overnight at 4°C with the following antibodies: Anti-GAPDH mouse monoclonal (cat. no. 60004-1-Ig; 1:5,000; Proteintech Group, Inc.), anti-microsomal glutathione S-transferase 1 (MGST1) rabbit monoclonal (cat. no. A0880; 1:500; ABclonal Biotech Co., Ltd.), anti-HK2 rabbit polyclonal (1:1,000; cat. no. A0994; ABclonal Biotech Co., Ltd,) and anti-Phosphatidylinositol-4-Phosphate 3-Kinase C2 Domain-Containing Subunit Gamma (PIK3C2G) polyclonal antibody (cat. no. 25904-1-AP; 1:2,000; Proteintech Group, Inc.). Following three washes with TBST (0.05% Tween-20), the membranes were incubated with horseradish peroxidase-conjugated secondary antibody (1:1,000; Beyotime Institute of Biotechnology, cat. no. A0208) at room temperature for 1 h. The membranes were washed three additional times with TBST, and chemiluminescent signals were detected using an ultra-sensitive ECL substrate (cat. no. BL520B, Biosharp, China) and imaged with a gel imaging system (Bio-Rad ChemiDoc XRS+; Bio-Rad Laboratories, Inc.). Data were analyzed using Image Lab 6.0 software (Bio-Rad), and semi-quantification of band intensities was performed with ImageJ 1.54d (National Institutes of Health, USA).

Echocardiography was performed using a Visualsonic Vevo 2100 (VisualSonics, Inc.) equipped with a 30 MHz linear array ultrasound transducer (20). Echocardiography was performed under inhalant anesthesia using isoflurane delivered via a nose cone connected to an anesthetic vaporizer. Induction was achieved with 4% isoflurane and maintenance was performed using 1.5–2.0% isoflurane. The internal diameter of the left ventricle (LV) was measured in M-mode over ≥3 cardiac cycles, and the mean value was subsequently calculated. The left ventricular end-diastolic inner diameter (LVEDd) and left ventricular end-systolic inner diameter (LVESd) were measured at the maximum and minimum LV areas, respectively. The left ventricular shortening fraction (LVFS) and left ventricular ejection fraction (LVEF) were calculated as follows: LVFS %=(LVEDd-LVESd)/LVEDd ×100%.

The myocardial tissues were fixed in 4% paraformaldehyde (PFA, cat. no. BL539A, Biosharp, China) at 4°C for 24–48 h, processed through graded ethanol, cleared in xylene, and embedded in paraffin. Samples were sliced into 5-µm-thick serial sections were prepared, dewaxed in xylene at room temperature (25°C) for 30 min and hydrated with graded alcohol. Antigens were retrieved using microwave heating at 95–100°C with sodium citrate buffer (pH 6.0; cat. no. 36311-B, Yeasen, China) for 30 min. Endogenous peroxidase activity was blocked with 3% hydrogen peroxide for 10 min at room temperature. Next, 5% bovine serum albumin (cat. no. HY-D0842; MedChemExpress) was applied at 37°C for 1 h. The sections were incubated overnight at 4°C with anti-MGST1 rabbit monoclonal (1:100) and anti-HK2 rabbit polyclonal antibody (1:200), followed by incubation with liquid 3,3′-diaminobenzidine (cat. no. 36311-H and 36311-I, Yeasen) substrate for color development according to the manufacturer's instructions. Sections were stained with hematoxylin at room temperature for 3 min to visualize nuclei, differentiated using 1% hydrochloric acid alcohol, dehydrated through ascending ethanol series (70, 95, and 100% ethanol; 2 min each at room temperature) and cleared in xylene (two changes, 3 min each at room temperature) prior to mounting with neutral balsam.

Stained sections were visualized and digitized using an Aperio CS2 digital slide scanner (Leica Biosystems, Germany) under bright-field microscopy.

Cardiac tissue was fixed in 4% paraformaldehyde (PFA, cat. no. BL520B, Biosharp, China) at 4°C for 24 h. Following fixation, the samples were washed in phosphate-buffered saline (PBS) and processed through a graded ethanol series (70, 85, 95, and 100%), cleared in xylene, and embedded in paraffin using an automated tissue processor (Leica TP1020; Leica Biosystems). Serial sections (5 µm thickness) were cut using a rotary microtome (Leica RM2235; Leica Biosystems) and mounted on poly-L-lysine-coated glass slides for subsequent histological staining. The sections were stained with H&E following deparaffinization and dehydration as previously described (21). Following deparaffinization and rehydration, the sections were stained with hematoxylin solution for 5 min, rinsed in running tap water for 5 min, differentiated in 1% acid alcohol for 30 sec, and blued in 0.2% ammonia water. Subsequently, the sections were counterstained with eosin for 3 min, dehydrated through a graded ethanol series (70, 95, and 100%), cleared in xylene, and mounted with neutral balsam. Additionally, sections were stained using a modified Masson's Trichrome Stain kit (cat. no. G1340; Beijing Solarbio Science & Technology Co., Ltd.) according to the manufacturer's instructions. Sections were visualized using an Aperio CS2 digital slide scanner (Leica Biosystems, Germany) under bright-field microscopy.

To investigate the regulatory effect of HK2 overexpression on downstream gene expression in HL-1 cells, cultured wild-type (WT) HL-1 cells (HK2-WT) and lentivirus-transfected HK2 overexpressing HL-1 cells (HK2-OE) were washed with a sterile PBS solution, lysed by adding TRIzol (cat. no. 15596026CN; Thermo Fisher Scientific, USA) lysis solution and incubated for 5 min at room temperature. The lysed cells were mixed and transferred to 1.5 ml sterile Eppendorf tubes for storage at −80°C. The samples were preserved and transported on dry ice to Metware Biotechnology Co., Ltd. (Wuhan, China) for analysis. Total RNA was processed with the NEBNext Ultra II RNA Library Prep kit (Cat. No. E7770; New England Biolabs) for Illumina sequencing. RNA integrity was verified using the Qsep400 Bioanalyzer (Bioptic Inc.) with all samples exhibiting RNA Integrity Number >7.0. Sequencing was performed on the Illumina NovaSeq 6000 platform using the NovaSeq 6000 SP Reagent Kit (300 cycles; Cat. No. 20027464; Illumina Inc.) for paired-end 150 bp sequencing. Final libraries were quantified by qPCR using the KAPA Library Quantification Kit (Cat. No. KK4824; Roche) and loaded at 12.5 pM concentration. Unsupervised principal component analysis) was performed by statistics function prcomp within R (version 3.5.1; r-project.org). The data was unit variance scaled before unsupervised PCA.

To investigate the effects of HK2 gene expression on energy and lipid metabolism in HL-1 cells, targeted quantification of metabolites was performed using liquid chromatography-tandem mass spectrometry (LC-MS/MS) as previously described (22). All metabolites were detected through MetWare (metware.cn/) using the AB Sciex QTRAP 6500 LC-MS/MS platform. The analysis was performed in both positive and negative ionization modes, with the following parameters: Ion spray voltage set at 5,500 V (positive) and −4,500 V (negative), curtain gas at 35 psi, and source temperature maintained at 550°C. Multiple reaction monitoring (MRM) transitions were optimized for each metabolite, with specific precursor and product ion pairs (m/z) selected for quantification. The nebulizer gas flow rate was maintained at 1.5 l/min, and the collision gas pressure was set to medium. Pearson correlation analysis was performed using quantitative values of genes and metabolites from all samples. Fold-change differences in correlation coefficients with absolute values >0.8 and P<0.05 were analyzed.

The cells were pelleted by centrifugation at 1,000 × g for 10 min at 4°C, and the resulting cell pellets were stored at −80°C until metabolomic analysis. The collected cell pellet samples for metabolomic analysis were thawed on ice and 100 µl ultrapure water was added to resuspend the cell pellet. A 50-µl aliquot of the cell suspension was added to 200 µl methanol (precooled to −20°C). The mixture was vortexed for 2 min at 1,000 × g at 4°C. The sample was flash-frozen in liquid nitrogen (−196°C) for 5 min, thawed on ice (0–4°C) for 5 min and vortexed at maximum speed (700 × g) at room temperature for 2 min. The freeze-thaw-vortex cycle was repeated three times. The sample was centrifuged at 12,000 × g for 10 min at 4°C and 200 µl supernatant was transferred to a new centrifuge tube and stored at −20°C for 30 min. The supernatant was centrifuged again at 12,000 × g for 10 min at 4°C. A total of 180 µl supernatant was filtered through a protein precipitation plate for LC-MS analysis, while the 50 µl cell suspension underwent three freeze-thaw cycles, followed by centrifugation at 12,000 × g for 10 min at 4°C. The protein concentration in the supernatant was determined using a BCA protein assay kit.

Total RNA was extracted from cells using TRIzol® reagent (Invitrogen; Thermo Fisher Scientific, Inc., Waltham, MA, USA) and reverse-transcribed into cDNA using HiScript IV RT SuperMix for qPCR (+gDNA wiper) (cat. no. R423-01; Vazyme Biotech Co., Ltd., Nanjing, China). Quantitative PCR was performed using the CFX Connect™ Real-Time PCR Detection System (Bio-Rad Laboratories, Inc., Hercules, CA, USA) with Taq Pro Universal SYBR qPCR Master Mix (cat. no. Q712; Vazyme Biotech Co., Ltd.). The thermal cycling conditions consisted of an initial denaturation at 95°C for 30 sec, followed by 40 cycles of 95°C for 10 sec and 60°C for 30 sec. Gene expression levels were normalized to GAPDH reference genes and calculated using the 2-ΔΔCq method (23). All primers were synthesized by Sangon Biotech Co., Ltd. (Shanghai, China; Table SI).

PPI networks were constructed using the STRING database (version 12.0; minimum required interaction score, 0.4; http://cn.string-db.org/) with the differentially expressed genes as input. The networks were visualized and analyzed using Cytoscape software (version 3.9.0; National Institute of General Medical Sciences), with the CytoHubba plugin (version 0.1, http://apps.cytoscape.org/apps/cytohubba) used to identify hub genes based on topological parameters (degree centrality and betweenness centrality).

All data are presented as the mean ± standard deviation from three independent experimental repeats. All analyses were conducted using GraphPad Prism 9.0 software (Dotmatics). Two samples were compared using unpaired two-tailed Student's t-test. For >2 groups, one-way ANOVA followed by Tukey's post hoc test was used if the data met the assumptions of normality (assessed by the Shapiro-Wilk test) and homogeneity of variance (verified by Levene's test). If these assumptions were violated, the non-parametric Kruskal-Wallis test, followed by Dunn's post hoc test, was applied. P<0.05 was considered to indicate a statistically significant difference.

To identify DEGs, a fold-change threshold of |log2 (fold-change)|≥1 and adjusted P<0.05 were employed. In the merged dataset of GSE4745 and GSE6880, 47 DEGs were identified, comprising 25 up- and 22 downregulated genes (Fig. 1A and B). To explore the biological functions of the DEGs and the potential biological pathways associated with DCM, GO functional and KEGG pathway enrichment analysis were performed (Tables SII and SIII).

GO enrichment analysis indicated DEGs were involved in BPs, including ‘fatty acid metabolic process’, ‘pyruvate metabolic process’, ‘small molecule catabolic process’, ‘response to toxic substance’, ‘response to starvation’, ‘response to hypoxia’ and ‘response to insulin’ (Fig. 1D). Additionally, the DEGs were localized to cellular components such as ‘banded collagen fibril’, ‘fibrillar collagen trimer’, ‘organelle outer membrane’, ‘sarcolemma’, ‘organelle inner membrane’, ‘mitochondrial inner membrane’ and ‘mitochondrial matrix’. Furthermore, DEGs associated with ‘hexose transmembrane transporter activity’, ‘glucose transmembrane transporter activity’, ‘neuropeptide hormone activity’, ‘CoA hydrolase activity’, ‘acyl-CoA hydrolase activity’, ‘glutathione peroxidase activity’, ‘platelet-derived growth factor binding’, ‘palmitoyl-CoA hydrolase activity’ and ‘phospholipid binding’ were significantly enriched (Fig. 1C-E).

KEGG enrichment analyses indicated that the DEGs were significantly enriched in ‘fatty acid elongation’, ‘drug metabolism-other enzymes’, ‘butanoate metabolism’, ‘arachidonic acid metabolism’, ‘metabolism of xenobiotics by cytochrome P450’, ‘drug metabolism-cytochrome P450’, ‘adipocytokine signaling pathway’, ‘glycolysis/gluconeogenesis’, ‘diabetic cardiomyopathy’, ‘carbon metabolism’, ‘linoleic acid metabolism’, ‘biosynthesis of amino acids’ and ‘HIF-1 signaling pathway’ (Fig. 1F and G). Additionally, GSEA demonstrated enrichment of DCM-associated genes in ‘collagen biosynthesis and modifying enzymes’ and ‘metabolism of lipids’ (Fig. 1H).

After conducting a WGCNA, the blue module was the most relevant to DCM (Fig. 2A-D). This module contained 19 functionally diverse hub genes that represent critical pathways in cardiac pathophysiology. The metabolic regulators included acyl-CoA thioesterases (ACOT1 and ACOT2) and acyl-CoA synthetase long-chain family member 6 (ACSL6), which coordinate lipid metabolism, along with β-hydroxybutyrate dehydrogenase (BDH1) and 3-hydroxy-3-methylglutaryl-CoA synthase 2 (HMGCS2), key enzymes in ketone body metabolism. Extracellular matrix remodeling was represented by collagen type I α2 chain (COL1A2) and collagen type III α1 chain (COL3A1), while immune modulation involved CD74 molecule (CD74) and cathepsin K. Metabolic enzymes included cytochrome P450 2E1 (CYP2E1), hexokinase 2 (HK2), and microsomal glutathione S-transferase 1 (MGST1). Additional significant genes were natriuretic peptide A, pyruvate dehydrogenase phosphatase catalytic subunit 2 (PDP2), proenkephalin (PENK), phosphatidylinositol-4-phosphate 3-kinase catalytic subunit type 2 gamma (PIK3C2G), phospholipase A2 group IIA (PLA2G2A), WAP four-disulfide core domain 1 (WFDC1), and the uncharacterized protein LOC102553868. These molecular networks provide crucial insights into the complex interplay of metabolic dysregulation, structural remodeling, and inflammatory responses in DCM development.

Using Venn analysis of DEGs and WGCNA module genes, 17 overlapping genes were identified (Fig. 3A). A LASSO regression model was constructed based on the expression matrix of these genes. Using the GLMNET package, the cross-validation curves (Fig. 3B) and LASSO coefficient path diagrams (Fig. 3C) of the model were visualized. Ultimately, four disease signature genes (HMGCS2, PIK3C2G, HK2 and MGST1) were identified. A comparative analysis of the expression of these genes was conducted in both the internal and external datasets using the limma package. Next, the ggpubr package was employed to visualize the differences in gene expression between the disease and control groups using boxplots (Fig. 3D and F). Compared with the control group, HK2 expression was downregulated in the DCM group, while the expression of HMGCS2, PIK3C2G and MGST1 was upregulated. Finally, the diagnostic value of signature genes for DCM was assessed using the pROC package. All four genes demonstrated high diagnostic value for DCM, with AUC >0.9 (Fig. 3E and G).

The surface area of HL-1 cardiomyocytes cultured in a high-glucose environment was notably increased compared with that of cardiomyocytes cultured in a low-glucose environment (Fig. 4A). A significant increase in cell surface area was confirmed by quantitative analysis following 48 h high-glucose treatment (Fig. 4B). The number of apoptotic cells was significantly higher following high-glucose compared with low-glucose treatment (Fig. 4D and E).

Western blot analysis revealed that MGST1 protein expression increased, while HK2 protein expression decreased in HL-1 cells treated with high- compared with low-glucose (Fig. 4E and F). PIK3C2G expression did not significantly differ between low-glucose and high-glucose treatment groups (Fig. 4E and F)

ROS production was significantly elevated in HL-1 cells treated with high-glucose, whereas ROS production was significantly decreased in HL-1 cells following HK2-OE (Fig. 4G and H). The increase in HK2 protein expression after HK2-OE was validated by a lentiviral transfection assay (Fig. 4I and J).

The present study found that HK2 was abundantly expressed in the nuclei of HL-1 cells cultured under high-glucose conditions (Fig. S1), suggesting it may influence the regulation of disease-associated genes in DCM. LLPS was detected in both HK2-WT and HK2-OE HL-1 cardiomyocytes by immunofluorescence analysis (Fig. S2).

In the present study, db/db mice were selected as a model for T2D, whereas db/m mice served as the control group. Western blot analysis of cardiac tissue revealed that HK2 was expressed at lower levels in the cardiac tissue of db/db compared with db/m mice. The difference in MGST1 and PIK3C2G cardiac tissue expression between the two groups was not significant (Fig. 4K and L). The heart-to-body weight ratio and the heart weight-to-tibia length ratio did not significantly differ between the two groups (Fig. 5A). Gross examination of the cardiac tissue indicated that the hearts of db/m mice were larger than those of db/db mice (Fig. 5B). A total of one db/db mouse died at ~32 weeks of age, and four of the surviving db/db mice exhibited significant frontal bladder overfill (4/5), suggesting the possibility of neurogenic cystitis with urinary retention (Fig. 5C).

To evaluate the effects of the intervention, blood glucose levels and body weight were measured at 8 and 32 weeks of age. At 8 weeks, the body weight of the db/db mice was significantly greater than that of the control mice, with a mean body weight ranging from 35 to 40 g. However, at 32 weeks, the db/db mice exhibited a significant decrease in body weight, and their mean weight was notably lower compared with that of the control mice (Fig. 5D). Blood glucose levels were significantly elevated in db/db mice compared with control mice at 8 and 32 weeks (Fig. 5E). Echocardiography data indicated that the LVEF and LVFS were significantly lower in db/db compared with db/m mice at 32 weeks (Fig. 5F). Additionally, the left intraventricular diameter was greater in db/db mice (Figs. 5G and S3), and diastolic function was impaired in these mice (Fig. 5H). Histological staining revealed increased cardiac fibrosis and cardiomyocyte hypertrophy in db/db mice (Fig. 5I and J).

Immunohistochemistry staining revealed no notable difference in MGST1 expression. By contrast with db/db mice, expression of HK2 was higher in the cardiac tissue of db/m mice. Similarly, the expression levels of HK2 in both db/m and db/db mice were greater than those in the PBS-treated controls (Fig. 5J).

Transcriptomic sequencing analysis was conducted on HK2-WT and HK2-OE cells. Energy and lipid metabolisms were quantitatively analyzed in six samples from each group. Transcriptomic sequencing was performed using Illumina high-throughput sequencing platform, which included three HK2-WT and three HK2-OE samples. Similarities between the samples were assessed through principal component analysis (PCA) (Fig. 6A). Gene expression was quantified using fragments per kilobase of transcript per million fragments mapped as a measure of gene expression levels (Fig. 6B). DEGs were identified with |log2 (fold-change)|≥1 and a false discovery rate <0.05.

A total of 1,292 DEGs were identified, comprising 432 up- and 860 downregulated genes (Fig. 6C). Energy metabolomics analysis was performed using a LC-MS/MS detection platform in conjunction with the MetWare database, resulting in the detection of 57 metabolites (Fig. 6D; Table SIV). Differences in metabolites within sample groups were analyzed using PCA (24) (Fig. 6E). In total, 11 differentially expressed metabolites were identified based on fold-change ≥2 and ≤0.5 (Fig. 6F and G; Table SV). Lipid metabolomics was used to quantify 1,269 lipids, and PCA was used to elucidate the overall metabolic differences and variability between samples within groups (Fig. 6H). Lipid cluster analysis revealed significant differential expression of genes associated with lipid metabolism (Fig. 6I). After screening for variable importance in projection value >1 and P<0.05, 471 differential lipids were identified (Fig. 6J; Table SVI).

KEGG pathway enrichment was compared to identify co-enriched KEGG pathways (Fig. 7A). DEGs and metabolites were enriched in ‘purine metabolism’, ‘nucleotide metabolism’ and ‘biosynthesis of cofactors’. KEGG enrichment analysis was conducted using the P-values from the transcriptome to identify common pathways within the top 50 most significantly enriched pathways. Transcriptomic analysis revealed significant enrichment of pathways associated with ‘glutathione metabolism’, ‘cholinergic synapse’, ‘HIF-1 signaling pathway’, ‘terpenoid backbone biosynthesis’, ‘insulin resistance’, ‘alcoholic liver disease’ and ‘purine metabolism’ (P<0.05). Additionally, metabolites involved in ‘nucleotide metabolism’ were significantly enriched (Fig. 7B). Both histologies exhibited significant enrichment in ‘purine metabolism’ (Fig. 7C).

Gene and metabolite expression patterns showed consistent differential regulation in quadrants 3 and 7, but opposing patterns in quadrants 1 and 9 (demarcated by black dashed lines), which visualizes gene-metabolite fold-change relationships 1 and 9 (Fig. 7D). DEGs and metabolites enriched in ‘purine metabolism’ were selected for correlation analysis. A total of four energy-associated metabolites were significantly associated with 14 genes involved in ‘purine metabolism’ (Fig. 7E). Finally, by constructing a correlation network of genes and metabolites, 114 genes associated with seven energy metabolites were identified (Fig. 7F).

Transcriptomics and lipid metabolomics of the co-enriched KEGG pathways were compared. Metabolites associated with ‘glycerolipid metabolism’, ‘lipid and atherosclerosis’, ‘sphingolipid signaling pathway’, ‘insulin resistance’, ‘ether lipid metabolism’ and ‘glycerophospholipid metabolism’ were enriched (Fig. 8A). Transcriptomic data revealed enrichment in ‘vascular smooth muscle contraction’, ‘Rap1 signaling pathway’, ‘human cytomegalovirus infection’, ‘cAMP signaling pathway’, ‘MAPK signaling pathway’, ‘amoebiasis’ and ‘calcium signaling pathway’ (Fig. 8A). Downstream genes of HK2, which are involved in various BPs, were significantly enriched in ‘calcium signaling pathway’, ‘thyroid hormone synthesis’, ‘vascular smooth muscle contraction’, ‘insulin secretion’, ‘MAPK signaling pathway’, ‘regulation of lipolysis in adipocytes’, ‘aldosterone synthesis and secretion’, ‘Rap1 signaling pathway’, ‘cAMP signaling pathway’, ‘inflammatory mediator regulation of TRP channels’ and ‘AGE-RAGE signaling pathway in diabetic complications’. Additionally, metabolites associated with ‘AGE-RAGE signaling pathway in diabetic complications’ and ‘sphingolipid signaling pathway’ were also significantly enriched (Fig. 8B and C). Numerous metabolites correlated with specific genes (Fig. 8D). Based on this analysis, regulatory network maps for genes and metabolites associated with ‘AGE-RAGE signaling pathway in diabetic complications’ and ‘sphingolipid signaling pathway’ were constructed (Fig. 8E and F).

A total of 14 genes [I16, I11A, thrombomodulin (THBD), angiotensinogen (AGT), COL1A1, COL3A1, endothelin-1 (EDN1), phospholipase C (Plc)b1 (PLCB1), COL4A2, CCL2, TGFB2, SERPINE1, PIK3CD and protein kinase Cβ (PRKCB)] and 32 lipid metabolites were identified by combined analysis of transcriptomics and lipid metabolomics within ‘AGE-RAGE signaling pathway in diabetic complications’ and significantly enriched (Fig. 8E). Of these, I16 and I11A were excluded due to the absence of specific primers (Table SI). Quantitative PCR analysis of the mRNAs of the remaining 12 genes in HL-1 cardiomyocytes from the WT and HK2-OE strains indicated that the difference between the subgroups was not significant for COL3A1. PRKCB exhibited high expression levels in HL-1 cardiomyocytes from the HK2-OE strain, while the other genes were expressed at low levels (Fig. 9A-C). Additionally, protein-protein interaction network maps were constructed for the remaining 12 genotypes (Fig. 9D).