Multi-omics profiling reveals coordinated metabolic and immuno-inflammatory remodeling in a high-fat diet and single-dose streptozotocin-induced mouse model of diabetic cardiomyopathy

Abstract

Background

Diabetic cardiomyopathy (DCM) is a major myocardial complication of type 2 diabetes and a contributor to heart failure. However, coordinated remodeling across the cardiac transcriptome, proteome and metabolome in DCM remains incompletely defined.

Methods

Type 2 diabetes was induced in mice by a high-fat diet combined with single-dose streptozotocin. DCM was characterized by echocardiography, histology, transmission electron microscopy, and systemic metabolic and inflammatory assessments. Left ventricular myocardium underwent RNA sequencing, quantitative proteomics and untargeted metabolomics, followed by integrated pathway and network analyses. Selected pathways were further examined by biochemical assays, with additional evaluation of mitochondrial injury and cGAS–STING signaling.

Results

Diabetic mice developed systolic and diastolic dysfunction with ventricular remodeling, interstitial fibrosis, mitochondrial ultrastructural damage, insulin resistance, and heightened systemic and myocardial inflammation. Transcriptomic and proteomic profiling revealed concordant suppression of sarcomeric organization, calcium handling, and mitochondrial oxidative metabolism, indicating downregulation of a contraction–calcium–mitochondrial energetics axis. Metabolomics showed reduced pyruvate-linked intermediates and enrichment of pentose phosphate, nucleotide and glutathione-related metabolites, consistent with diversion of central carbon flux from mitochondrial oxidation toward redox and biosynthetic programs. Across all three omics layers, phospholipid–choline metabolism emerged as a remodeled hub connecting central carbon pathways with complement, cell adhesion and inflammatory signaling. Increased cGAS–STING activity in DCM myocardium further linked mitochondrial injury to innate immune activation.

Conclusions

Integrated multi-omics analysis reveals coordinated remodeling across mitochondrial function, metabolism, inflammation, and fibrosis in experimental DCM. Alterations in phospholipid–choline metabolism, pentose phosphate and central carbon pathways, cGAS–STING signaling, and the contraction–calcium–mitochondrial module were concurrently observed and collectively characterize the molecular landscape of diabetic cardiomyopathy. These findings provide a framework for future mechanistic studies and may help identify candidate biomarkers and potential therapeutic pathways for further investigation.

Graphical Abstract

Integrated multi-omics remodeling in HFD/STZ-induced diabetic cardiomyopathy. HFD plus single-dose STZ induced DCM with metabolic dysfunction, inflammation, mitochondrial injury, cardiac dysfunction, and fibrosis. Integrated transcriptomic, proteomic, and metabolomic analyses revealed suppression of the contraction–calcium–mitochondrial energetics axis, central carbon reprogramming toward the pentose phosphate pathway, phospholipid–choline remodeling, and cGAS–STING-associated innate immune and fibrotic activation. Green arrows indicate activation, red dashed lines indicate suppression, and gray lines indicate associations.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12872-026-06682-z.

Introduction

Diabetic cardiomyopathy (DCM) is a distinct myocardial disorder in people with diabetes, characterized by structural and functional cardiac abnormalities in the absence of coronary artery disease, hypertension, valvular heart disease, or other established causes of cardiomyopathy [1–3]. DCM commonly begins with diastolic dysfunction and subclinical ventricular remodeling and may progress to systolic dysfunction and heart failure [1, 4]. With the increasing prevalence of type 2 diabetes mellitus (T2DM), DCM represents an important and often underrecognized cardiovascular complication [5–8]. Its pathogenesis is multifactorial and involves chronic hyperglycemia and insulin resistance, altered substrate utilization, mitochondrial dysfunction and oxidative stress, impaired calcium handling, immune-inflammatory activation, and progressive extracellular matrix remodeling [9–11].

Despite extensive investigation, most previous studies have focused on individual pathways or single molecular layers. Consequently, how changes in gene expression, protein abundance, and metabolite profiles interact to drive the transition from metabolic stress to inflammation, fibrosis, and cardiac dysfunction remains incompletely understood. Integrated multi-omics approaches provide an opportunity to identify coordinated molecular programs that cannot be fully captured by single-omics analyses [12–15]. Recent studies have applied combinations of transcriptomics, proteomics, metabolomics, and other omics approaches to diabetic heart disease and have revealed important alterations in lipid metabolism, oxidative stress, and myocardial energy metabolism [16–18]. However, integrated multi-layer characterization of diabetic cardiomyopathy induced by a high-fat diet (HFD) combined with single-dose streptozotocin (STZ) remains limited, and the coordinated molecular remodeling underlying this model has not been fully defined.

In the present study, we used an HFD plus single-dose STZ mouse model of T2DM-associated cardiomyopathy [19–22] and integrated myocardial transcriptomic, quantitative proteomic, and untargeted metabolomic profiling with comprehensive cardiac phenotyping. We aimed to characterize coordinated molecular alterations across multiple biological layers and to identify major pathways linking metabolic and mitochondrial dysfunction with immune-inflammatory and fibrotic remodeling. These analyses were further complemented by targeted molecular validation to nominate candidate pathways and molecular nodes for future mechanistic investigation.

Materials and methods

Animals and study design

All animal procedures were conducted in accordance with institutional guidelines for the care and use of laboratory animals and were approved by the Animal Ethics Committee of Nankai University. Mice were housed in a specific pathogen-free (SPF) facility under controlled environmental conditions (20–24℃, 30–70% relative humidity, 12-h light–dark cycle). Animals had ad libitum access to food and water throughout the study, including an acclimatization period of at least 1 week.

Eight-week-old male C57BL/6J mice were purchased from Weitong Lihua Experimental Animal Technology Co., Ltd. (Beijing, China) and randomly assigned to a DCM group or a control (CT) group (n = 6 per group). The DCM group was fed a commercially available HFD, whereas control mice received standard chow. The detailed composition of both diets is provided in Supplementary Table S1. After 6 weeks of HFD or chow feeding, intraperitoneal glucose tolerance tests (IPGTT) and intraperitoneal insulin tolerance tests (IPITT) were performed to assess the development of glucose intolerance and insulin resistance. In week 7, mice in the DCM group were fasted for 12 h and received a single intraperitoneal injection of STZ (75 mg/kg, S0130, Sigma, Livonia, MI, USA) freshly dissolved in 0.1 mol/L citrate buffer (pH 4.5) at a concentration of 10 mg/mL. Control mice received vehicle alone.

All animals were maintained on their respective diets until week 22. Subsequently, they conducted IPGTT and IPITT again. At study completion, transthoracic echocardiography was performed to assess cardiac structure and function. Mice were then anesthetized and euthanized by carbon dioxide asphyxiation, and blood and left ventricular myocardium were collected for subsequent biochemical, histological, and multi-omics analyses.

IPGTT and IPITT

IPGTT and IPITT were performed after a 6 h–4 h fast, respectively, with free access to water. For each test, 6 control mice and 6 diabetic mice were examined. After measurement of baseline tail-vein glucose (0 min), mice undergoing IPGTT received an intraperitoneal injection of D-glucose (1.5 g/kg body weight, dissolved in sterile saline). For IPITT, mice received an intraperitoneal injection of insulin (0.75 U/kg body weight, diluted in sterile saline). Tail-vein glucose was measured at 15, 30, 60, and 120 min after injection using a handheld glucometer. All tests were conducted at constant room temperature. During IPITT, if glucose fell below 2.8 mmol/L, D-glucose solution was immediately administered for rescue. IPGTT and IPITT data were analyzed using two-way repeated-measures ANOVA with group and time as factors, followed by appropriate multiple-comparisons testing. Areas under the curve (AUCs) were calculated for individual animals and compared between two groups using an unpaired two-tailed Student’s t-test.

Treadmill exercise endurance test

Exercise endurance was assessed using a motorized treadmill test to evaluate functional capacity. All mice were acclimatized to the treadmill for 3 consecutive days before the formal assessment (10 min/day at 10 m/min). During the test, mice ran on an electrified treadmill at a level grade. The initial speed was set at 8 m/min and increased by 2 m/min every 3 min until reaching a maximum speed of 26 m/min. Exhaustion was defined as the inability to resume running after remaining on the shock grid for 10 consecutive seconds. Total running distance to exhaustion was recorded as the primary index of exercise endurance.

Echocardiography

Transthoracic echocardiography was performed using a high-resolution small-animal imaging system (Vevo 2100, VisualSonics, Toronto, Canada) equipped with an MS400 transducer. Mice were anesthetized with 1.5% to 2.0% isoflurane and placed in the supine position on a temperature-controlled platform to optimize the left parasternal acoustic window. Heart rate was maintained between 450 and 550 bpm during image acquisition. Two-dimensional and M-mode images were acquired from parasternal short-axis views at the mid-papillary level. Left ventricular internal dimension at end-diastole (LVIDd) and end-systole (LVIDs) were measured, and left ventricular ejection fraction (LVEF) and fractional shortening (FS) were calculated to assess systolic function. Diastolic function was evaluated using pulsed-wave Doppler in the apical four-chamber view to record transmitral inflow and to measure early (E) and late (A) peak diastolic velocities, from which the E/A ratio was derived. All parameters were averaged over three consecutive cardiac cycles. Echocardiographic acquisition and analysis were performed in accordance with recommendations from the American Society of Echocardiography.

Histological assessment of myocardial remodeling

Left ventricular tissue was fixed in 10% phosphate-buffered formalin and processed using standard histological procedures. Transverse sections (5 μm) were obtained at the mid-ventricular level and stained as described below. Hematoxylin and eosin (H&E) staining was used to assess overall myocardial morphology. Wheat germ agglutinin (WGA) staining was performed to evaluate cardiomyocyte hypertrophy. Cardiomyocyte cross-sectional area was quantified in high-power fields (400×) using a light microscope (Olympus) and ImageJ software (National Institutes of Health, USA). For each heart, 5 to 8 non-overlapping fields were analyzed, and at least 80 to 100 cardiomyocytes were measured.

Masson’s trichrome staining was used to assess myocardial fibrosis. Interstitial collagen content was quantified as the percentage of collagen-stained area relative to total myocardial area in 5 to 8 fields per section. For ultrastructural assessment, small blocks of left ventricular tissue were fixed in 2.5% glutaraldehyde, post-fixed in osmium tetroxide, embedded in epoxy resin, and cut into ultrathin sections. Sections were stained with uranyl acetate and lead citrate and examined by transmission electron microscopy to evaluate mitochondrial ultrastructure in cardiomyocytes. All histological and morphometric analyses were performed by an investigator blinded to group allocation.

Quantitative real-time polymerase chain reaction

Total RNA was extracted from left ventricular myocardium using TRIzol reagent (15596026CN, Thermo Fisher Scientific, USA) according to the manufacturer’s instructions. RNA concentration and purity were determined spectrophotometrically. Equal amounts of RNA were reverse-transcribed into complementary DNA using the PrimeScript RT reagent kit (AG0304, SparkJade, Shandong, China). Quantitative real-time PCR was performed with SYBR Green chemistry (AH0104, SparkJade, Shandong, China) on a StepOnePlus Real-Time PCR System (Applied Biosystems, USA). Relative mRNA expression of target genes was normalized to the geometric mean of beta-actin (ACTB) and hypoxanthine phosphoribosyltransferase 1 (HPRT1) and calculated using the 2^−ΔΔCt method. Primer sequences are provided in Supplementary Table S2.

Biochemical assays of serum cardiac injury markers and myocardial ATP

Serum lactate dehydrogenase (LDH), creatine kinase-MB (CK-MB), and lactate (LAC) were quantified using enzymatic colorimetric assay kits (P0393S, S0287S, and S0204S, Beyotime, Shanghai, China) according to the manufacturers’ instructions. Approximately 20µL of serum from each mouse was used for each biochemical assay. Myocardial adenosine triphosphate (ATP) levels were measured using a commercial ATP assay kit (S0026, Beyotime, Shanghai, China). Briefly, approximately 20 mg of left ventricular myocardium from each mouse was homogenized according to the manufacturer’s protocol, and ATP levels in the clarified supernatant were quantified using a luciferin-luciferase-based assay. Total protein concentration in the same myocardial extracts was determined using a bicinchoninic acid (BCA) protein assay kit (PC0020, Solarbio, Beijing, China). ATP content was normalized to the total protein amount and expressed as µmol/g protein.

Enzyme-linked immunosorbent assay

Concentrations of selected molecules in mouse serum and left ventricular homogenates were quantified using commercial enzyme-linked immunosorbent assay kits according to the manufacturers’ instructions. Pro-inflammatory cytokines, including interleukin 1β (IL-1β, ab197742, Abcam, Cambridge, UK), tumor necrosis factor α (TNF-α, ab208348, Abcam), interleukin 6 (IL-6, ab178013, Abcam), and interleukin 18 (IL-18, kit information as specified by the manufacturer), as well as the anti-inflammatory cytokine interleukin 10 (IL-10; ab255729, Abcam), were measured to assess inflammatory status. Markers of cardiac strain and pyroptosis, including N-terminal pro-B-type natriuretic peptide (NT-proBNP, ml324452, Mlbio, Shanghai, China) and caspase 1 (ml107151,Mlbio, Shanghai, China), were measured in serum or myocardial homogenates. In addition, serum pyruvate (MP0751, NeoScientific, Cambridge, MA, USA) and acetylcholine (ACh, ml093045, Mlbio, Shanghai, China) levels were determined using competitive enzyme-linked immunosorbent assay kits. All assays were performed following the respective kit protocols, and each sample was measured in duplicate.

Transcriptomics analysis

Three mice per group were randomly selected for transcriptomic analysis, and approximately 50 mg of left ventricular myocardium from each individual mouse was used for total RNA extraction. using TRIzol reagent (15596026CN, Thermo Fisher Scientific, USA) according to the manufacturer’s instructions. RNA concentration and purity were assessed using a NanoDrop ND-1000 spectrophotometer (9053360, Thermo Fisher Scientific, Waltham, MA, USA), and RNA integrity was evaluated using an Agilent 2100 Bioanalyzer (G2939A, Agilent Technologies, USA). RNA sequencing libraries were prepared and sequenced on an Illumina NovaSeq 6000 platform to generate 150-bp paired-end reads.

Raw reads were trimmed and quality-filtered to remove adaptor sequences and low-quality bases. Reads containing more than 5% ambiguous nucleotides (N) were excluded. Clean reads were aligned to the mouse reference genome using HISAT2 (v2.2.1, Johns Hopkins University, USA), and gene-level read counts were generated using featureCounts. Differential expression analysis was performed using DESeq2 (v1.22.2, Bioconductor, USA). Genes with an absolute log₂ fold change (|log₂FC|) > 1 and a Benjamini–Hochberg adjusted p value < 0.05 were considered differentially expressed. Normalized expression values were used for unsupervised hierarchical clustering and heatmap visualization, and volcano plots were generated to summarize differential expression patterns. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were performed separately for up-regulated and down-regulated genes using R-based packages, and results were visualized as bar plots and bubble plots.

Proteomics analysis

Three mice per group were randomly selected for quantitative proteomic analysis, and 30 mg left ventricular myocardium from each individual mouse was used for protein extraction. Total protein was extracted using SDS lysis followed by cold acetone precipitation. Briefly, tissue was homogenized in ice-cold lysis buffer (8 M urea, 2% SDS, 1× protease inhibitor cocktail) by vortexing and high-throughput mechanical grinding (three cycles). Homogenates were centrifuged at 12,000×g for 20 min at 4 °C, and supernatants were collected. Protein concentration was determined using a BCA protein assay kit. Equal amounts of protein were reduced, alkylated, and digested with sequencing-grade trypsin (V5111, Promega, WI, USA). Peptides were desalted, reconstituted in 0.1% trifluoroacetic acid, and fractionated by high-pH reversed-phase chromatography using a Pierce High pH Reversed-Phase Peptide Fractionation Kit (84868, Thermo Fisher Scientific, USA). Eight primary fractions were pooled into six fractions and dried under vacuum.

For LC–MS/MS analysis, peptides were separated on an EASY-nLC 1200 system (LC140, Thermo Fisher Scientific, USA) coupled to an Orbitrap Exploris 480 mass spectrometer (BRE725539, Thermo Fisher Scientific, USA) operated in data-independent acquisition mode. Raw DIA data were processed using Spectronaut (version 13, Biognosys, Switzerland) with default settings, and label-free quantification was performed based on normalized peptide intensities. Functional annotation of identified proteins was performed using Gene Ontology, Kyoto Encyclopedia of Genes and Genomes, and the Clusters of Orthologous Groups and euKaryotic Orthologous Groups databases. Differentially expressed proteins (DEPs) were defined as those with |log2 fold change| > 0.58 (fold change > 1.5) and q value < 0.05 after multiple-testing correction.

Metabolomics analysis

Five mice per group were randomly selected for metabolomic analysis, and approximately 25 mg of left ventricular myocardium from each individual mouse was homogenized in 500 µL of extraction solvent containing internal standards (methanol: acetonitrile = 2:2:1, v/v/v). Samples were vortexed, mechanically ground (three cycles; 4 min at 35 Hz), and sonicated in an ice-water bath for 5 min. Extracts were incubated at − 40 °C for 1 h and centrifuged at 12,000 rpm for 15 min at 4 °C. Supernatants were transferred to vials for LC–MS/MS analysis. A pooled quality control sample was generated by combining equal aliquots of supernatant from each sample.

Untargeted metabolomics was performed on a Vanquish UHPLC system (IQLAAAGABHFAPUMBJC, Thermo Fisher Scientific, USA) equipped with a Phenomenex Kinetex C18 column (2.1 × 50 mm, 2.6 μm) and coupled to an Orbitrap Exploris 120 mass spectrometer (Thermo Fisher Scientific, USA). Raw data were converted to mzXML using ProteoWizard and processed using XCMS within BiotreeDB (v2.1, Biotree Biomedical Technology, China) for peak detection, retention time alignment, and peak integration. Metabolites were annotated by matching experimental spectra against an in-house MS/MS library implemented in BiotreeDB (match score cutoff = 0.3). The resulting normalized peak area matrix was used for downstream analyses.

Multivariate analyses were performed in SIMCA (v16.0.2, Umetrics, Sweden), including principal component analysis for overview separation and data quality assessment, and orthogonal projections to latent structures–discriminant analysis for feature selection. Model validity was evaluated by 7-fold cross-validation and 200 permutation tests. Differential metabolites were defined as features with variable importance in projection > 1.0 and Benjamini–Hochberg adjusted P value < 0.05 (Student’s t-test). Volcano plots and hierarchical clustering heatmaps were used for visualization. Pathway enrichment and topology analyses were performed using KEGG and MetaboAnalyst, and pooled quality control samples were injected regularly throughout the sequence to monitor system stability.

Integration of transcriptomic and proteomic data

For transcriptome-proteome integration, differentially expressed genes (DEGs) and DEPs were first defined as described above. Gene symbols were used to match mRNAs and proteins derived from the same locus. Three subsets were generated: (i) concordant molecules that were significantly regulated at both the mRNA and protein levels, (ii) mRNA-only changes that were significantly altered at the transcript level but not at the protein level and (iii) protein-only changes that were significantly altered at the protein level but not at the transcript level. For the concordant subset, normalized transcript and protein abundances were used to calculate Spearman correlation coefficients across samples, and the overall mRNA–protein correlation was summarized.

Functional enrichment of the concordant set was performed using GO and KEGG analyses implemented in R (clusterProfiler and related packages). Enriched GO terms and KEGG pathways were visualized as bar plots, highlighting processes related to sarcomere structure, calcium handling, energy metabolism and inflammatory signaling. For the discordant subsets (mRNA-only and protein-only), GO/KEGG enrichment was also performed to characterize pathways showing predominant post-transcriptional regulation. Heatmaps and circular plots were generated from normalized z-scores to display the direction and magnitude of regulation across the two omics layers.

Integration of transcriptomic, proteomic and metabolomic data

Tri-omics integration was performed at pathway and network levels by combining differentially expressed genes, proteins, and metabolites. KEGG pathway enrichment was conducted separately for each omics layer, and pathways significantly enriched (adjusted P value < 0.05) in at least two layers were defined as shared multi-omics pathways. For each shared pathway, the numbers of contributing genes, proteins, and metabolites were summarized and visualized using combined plots.

To project coordinated changes onto representative metabolic routes, selected KEGG pathways were visualized using the pathview package in R. Log2 fold changes of genes and proteins were mapped to enzyme nodes, whereas log2 fold changes of metabolites were mapped to metabolite nodes.

Correlation-based integration was performed to evaluate relationships between metabolite classes and gene or protein modules. Normalized abundance matrices of selected genes, proteins, and significantly altered metabolites were combined, and pairwise Spearman correlation coefficients were calculated. P values were adjusted using the Benjamini–Hochberg method, and significant associations were defined as |r| ≥ 0.6 with adjusted P value < 0.05. The resulting correlation matrix was visualized as a clustered heatmap, with metabolite classes as rows and gene or protein modules as columns.

Finally, a multi-omics correlation network was constructed using significantly associated genes, proteins, and metabolites as nodes and significant correlations as edges. Nodes were annotated by omics layer and metabolite class, and community detection was performed using igraph, with network visualization in Cytoscape.

Data analysis and statistical analysis

For non-omics experiments, data are presented as mean ± SD. Statistical analyses were performed using GraphPad Prism (version 9.0, GraphPad Software, USA). Comparisons between two groups were conducted using unpaired two-tailed Student’s t-test. Time-course data for intraperitoneal glucose tolerance tests and intraperitoneal insulin tolerance tests were analyzed using two-way repeated-measures analysis of variance with group and time as factors, followed by Bonferroni post hoc tests. A P value < 0.05 was considered statistically significant.

For omics and integrated multi-omics analyses, statistical procedures and multiple-testing corrections are described in the corresponding subsections. Unless otherwise specified, significance was defined as adjusted P value (q value) < 0.05.

Results

Establishment and phenotypic validation of an HFD + STZ-induced DCM mouse model

As shown in the experimental workflow (Fig. S1), mice received a HFD feeding combined with STZ administration and were assessed after 22 weeks. Echocardiography indicated a robust DCM phenotype, characterized by left ventricular systolic dysfunction reflected by significantly reduced LVEF and LVFS, together with increased left ventricular (LV) internal diameters at end-diastole and end-systole, consistent with impaired contractility accompanied by chamber dilation (Fig. 1A, C–F). Transmitral Doppler further showed a reduced E/A ratio, supporting abnormal diastolic filling in DCM hearts (Fig. 1B, G).

Fig. 1.

Cardiac dysfunction and remodeling in an HFD + STZ-induced DCM mouse model. A Representative M-mode echocardiography images from CT and DCM mice. B Representative pulsed-wave Doppler recordings of transmitral inflow. C–F Quantification of left ventricular ejection fraction (LVEF), fractional shortening (LVFS), left ventricular internal diameter at end-diastole (LVIDd) and end-systole (LVIDs). G Mitral inflow E/A ratio. H Serum NT-proBNP levels. I Myocardial NT-proBNP content normalized to total protein. J Treadmill exercise distance. K Representative H&E-stained whole-heart cross-sections and quantification of right ventricular free-wall thickness. L Wheat germ agglutinin (WGA, green) staining with DAPI (blue) and quantification of left ventricular cardiomyocyte cross-sectional area. M Masson’s trichrome staining and quantification of myocardial fibrosis. N Representative transmission electron microscopy (TEM) images showing mitochondrial ultrastructure; boxed regions are shown at higher magnification. Data are shown as mean ± SD. Each dot represents one mouse with n = 5–6 per group. Statistical analysis was performed using an unpaired two-tailed Student’s t-test. **p < 0.01, ***p < 0.001, ****p < 0.0001. ns indicates not significant.

Concordant with functional deterioration, serum and myocardial NT-proBNP levels were markedly elevated, and treadmill testing revealed substantially reduced exercise capacity in DCM mice (Fig. 1H–J). Histological analyses showed cardiomyocyte hypertrophy, evidenced by increased WGA-defined cross-sectional area, and enhanced interstitial fibrosis by Masson’s trichrome staining, whereas right ventricular free-wall thickness remained unchanged (Fig. 1K–M). At the ultrastructural level, transmission electron microscopy revealed swollen mitochondria with disrupted cristae in DCM myocardium (Fig. 1N). Collectively, these results confirm successful establishment of an HFD + STZ-induced DCM model and highlight mitochondrial ultrastructural injury as a key feature for subsequent mechanistic investigations.

Impaired glucose homeostasis and inflammatory activation in HFD + STZ-induced DCM

We next assessed metabolic homeostasis and inflammatory status in the HFD + STZ-induced DCM model. At week 22, DCM mice exhibited impaired glucose tolerance during the IPGTT, characterized by greater glucose excursions and delayed glucose clearance compared with control mice. Consistently, the corresponding IPGTT AUC was significantly increased in the DCM group. During the IPITT, the glucose-lowering response to insulin was attenuated in DCM mice, accompanied by a significant difference in AUC, indicating reduced insulin sensitivity (Fig. 2A-B). In contrast, the week-6 pre-STZ IPGTT and IPITT data are provided in the Fig. S5 to illustrate the metabolic status of the animals before STZ administration. In line with these findings, serum insulin levels were significantly decreased, whereas blood glucose levels were increased in DCM mice (Fig. 2C–D).Serum CK-MB and LDH levels were elevated in DCM mice, consistent with aggravated cardiac injury (Fig. 2E–F). Myocardial ATP content was increased, while serum lactate remained unchanged (Fig. 2G–H), suggesting a shift in myocardial energetic status without systemic lactate accumulation.

Fig. 2.

HFD + STZ-induced DCM is associated with impaired glucose homeostasis and heightened inflammation. A Intraperitoneal glucose tolerance test (IPGTT) performed at week 22 of model induction: (a) blood glucose levels at 0, 15, 30, 60, and 120 min after glucose administration; (b) corresponding area under the curve (AUC). B Intraperitoneal insulin tolerance test (IPITT) performed at week 22 of model induction: (a) blood glucose levels at 0, 15, 30, 60, and 120 min after insulin administration; (b) corresponding AUC. CT, control; DCM, diabetic cardiomyopathy. C Serum insulin levels. D Blood glucose (GLU). E–F Serum myocardial injury markers creatine kinase-MB (CK-MB) (E) and lactate dehydrogenase (LDH). G Myocardial ATP content normalized to total protein. H Serum lactate (LAC). I–N Cytokine profiles in serum and left ventricular (LV) tissue, including IL-6 (I), IL-18 (J), IL-10 (K), IL-1β (L), TNF-α (M), and caspase-1 (N). Cytokine contents in LV tissue were normalized to total protein. Data are presented as mean ± SD. Each dot represents one mouse, with sample size indicated by the number of dots. Statistical analysis was performed using an unpaired two-tailed Student’s t-test. **p < 0.01, ***p < 0.001, ****p < 0.0001. ns indicates not significant.

Inflammatory profiling showed a marked pro-inflammatory phenotype in both the circulation and LV tissue. Levels of IL-6, IL-18, IL-1β and TNF-α were increased in serum and LV myocardium from DCM mice (Fig. 2I–J, L–M), whereas the anti-inflammatory cytokine IL-10 was reduced in both compartments (Fig. 2K). In parallel, caspase-1 levels were elevated in serum and LV tissue, supporting inflammasome-associated inflammatory activation (Fig. 2N). Collectively, these data indicate that HFD + STZ-induced DCM is accompanied by glucose intolerance and insulin resistance, together with heightened systemic and myocardial inflammation and inflammasome activation.

RNA-seq identifies coordinated transcriptional reprogramming in DCM myocardium

To characterize the transcriptional profile associated with the DCM phenotype, RNA sequencing was performed on LV myocardial tissue. Using |log2FC| ≥ 1 and FDR < 0.05 (n = 3 per group), we identified 476 DEGs in DCM versus CT hearts, including 150 up-regulated and 326 down-regulated transcripts (Fig. 3B). Unsupervised hierarchical clustering based on these DEGs distinctly separated DCM from CT samples, indicating good within-group consistency and substantial transcriptional remodeling in DCM myocardium (Fig. 3A).

Fig. 3.

Transcriptomic profiling identifies coordinated remodeling in DCM myocardium. A Heatmap of differentially expressed genes (DEGs) showing unsupervised hierarchical clustering of CT and DCM left ventricular (LV) samples. (B) Volcano plot of DEGs in DCM versus CT, with selected representative genes labeled. C Gene Ontology (GO) enrichment analysis of down-regulated DEGs, grouped by molecular function (MF), cellular component (CC) and biological process (BP). D GO enrichment analysis of up-regulated DEGs. E KEGG pathway enrichment for down-regulated DEGs (top 20). F KEGG pathway enrichment for up-regulated DEGs (top 20). DEGs were defined using the indicated thresholds of |log2FC| ≥ 1 and FDR < 0.05. Enrichment analyses were performed using over-representation analysis. The x-axis indicates –log10(P value), and bars denote the number of DEGs annotated to each term or pathway

Functional enrichment analysis revealed divergent signatures between down- and up-regulated programs. Down-regulated genes were significantly enriched in components and processes related to cellular membranes and junctions, particularly transmembrane transport, with a strong emphasis on ion transport and calcium ion handling (e.g., voltage-gated ion channel activity and calcium ion transmembrane transporter activity) (Fig. 3C). Consistently, KEGG pathway analysis highlighted alterations in excitation-contraction coupling and second-messenger signaling, including calcium signaling pathway, cAMP signaling pathway, and cGMP-PKG pathway. Cardiomyopathy-related pathways (hypertrophic, dilated, and arrhythmogenic right ventricular cardiomyopathy(ARVC)) and circadian entrainment were also enriched (Fig. 3E). These transcriptional changes align with the impaired systolic function and diastolic abnormalities observed in the DCM model.

In contrast, up-regulated DEGs were strongly enriched in immune- and inflammation-related programs. GO terms were dominated by innate immune response and defense response, alongside membrane- and receptor-associated components and activities (Fig. 3D). KEGG analysis further indicated heightened inflammatory and pattern-recognition signaling, including cell adhesion molecules, NOD-like receptor (NLR) signaling pathway, phagosome, and cytosolic DNA-sensing pathways (cGAS-STING), as well as pathways linked to immune activation (e.g., NK cell cytotoxicity) (Fig. 3F). Notably, the concomitant enrichment of inflammasome-related and cGAS-STING pathways corresponds with the elevated inflammatory cytokines and increased caspase-1 activity detected in serum and myocardium, suggesting coordinated inflammatory activation in DCM.

Proteomic profiling reveals enhanced immune and adhesion pathways with metabolic suppression in DCM myocardium

To profile proteomic alterations in DCM, quantitative proteomics was conducted on LV tissue. Compared with CT, 103 DEPs were identified in DCM left ventricle (35 up-regulated and 68 down-regulated) (Fig. 4B). DEPs were defined as FC ≥ 1.5 or ≤ 0.67 with p < 0.05. Unsupervised clustering based on these DEPs distinctly separated DCM from CT samples, indicating consistent proteomic alterations across biological replicates (Fig. 4A). Subcellular annotation revealed that DEPs were broadly distributed across cellular compartments, with the majority localized to the nucleus (57) and cytoplasm (50), followed by the extracellular space (45) and plasma membrane (24) (Fig. 4C). Fewer DEPs mapped to mitochondria (9) or lysosomes (4).

Fig. 4.

Proteomic profiling reveals immune and adhesion activation with metabolic suppression in DCM myocardium. A Heatmap of differentially expressed proteins (DEPs) showing unsupervised hierarchical clustering of left ventricular samples from CT and DCM mice. B Volcano plot of protein abundance changes in DCM versus CT, with selected representative proteins labeled. The absolute fold-change threshold is indicated (FC = 1.5). C Subcellular localization distribution of DEPs. D Gene Ontology (GO) enrichment analysis of DEPs, grouped by molecular function (MF), cellular component (CC) and biological process (BP). E KEGG pathway enrichment showing representative pathways for up-regulated (red) and down-regulated (blue) DEPs. F KEGG pathway enrichment for DEPs (top 20), displayed as a dot plot with rich factor and protein counts. G Protein–protein interaction (PPI) network of DEPs highlighting major functional modules. For enrichment plots, the x-axis represents –log10(P value). Bars or dots indicate the number or proportion of proteins annotated to each term or pathway.

Functional enrichment analysis highlighted an immune- and adhesion-dominant signature in the DCM proteome. Among DEPs, GO analysis indicated enrichment in catalytic and structural molecule activities, as well as in biological processes responsive to stimuli, including numerous immune system-related terms (Fig. 4D). Correspondingly, KEGG pathway analysis showed preferential enrichment of inflammatory and immune pathways among up-regulated proteins, including apoptosis and TNF signaling pathway, as well as pathways related to immune activation and cell-cell interactions (e.g., T cell receptor signaling pathway, complement and coagulation cascades, adherens junction, and cell adhesion molecules) (Fig. 4E, F). Pathways related to vascular smooth muscle contraction and the biosynthesis of unsaturated fatty acids were also enriched among up-regulated proteins (Fig. 4E).

Conversely, down-regulated DEPs were primarily associated with metabolic–detoxification pathways, including retinol metabolism and cytochrome P450-mediated drug and xenobiotic metabolism, together with glycosaminoglycan biosynthesis and cytoskeletal regulation (Fig. 4E). Network analysis further grouped DEPs into interconnected modules, wherein immune and complement-related and adhesion-related proteins constituted central hubs linking cytoskeletal and metabolic–detoxification modules (Fig. 4G). Taken together, these findings suggest that DCM myocardium exhibits coordinated proteome remodeling characterized by activation of immune and adhesion programs concurrent with the suppression of metabolic–detoxification and cytoskeletal regulatory pathways.

Integrated transcriptome–proteome analysis reveals limited mRNA-protein concordance and highlights contractile, Ca²⁺ handling and metabolic modules in DCM myocardium

To elucidate the relationship between transcriptional and proteomic remodeling in DCM myocardium, quantified proteins were matched to their corresponding transcripts. The matched pairs were stratified into three categories: molecules with concordant regulation (significant changes at both mRNA and protein levels), transcript-only changes, and protein-only changes (Fig. 5A). Notably, only 14 molecules were significantly altered at both the mRNA and protein levels, whereas most changes were detected in only one layer, indicating prevalent layer-specific regulatory control in DCM. Across all matched gene–protein pairs, the overall mRNA–protein association was modest but positive (Spearman R = 0.1392), with most pairs clustering near the origin in a fold-change comparison plot (Fig. 5B–C). Quadrant analysis further demonstrated that concordant directional changes between transcripts and proteins were relatively infrequent, whereas significantly oppositely regulated pairs were rare (Fig. 5F), supporting a prominent contribution of post-transcriptional regulation and altered protein turnover to the proteomic remodeling observed in DCM hearts. The Z-scored heatmap shows broadly concordant abundance patterns between the proteome and transcriptome across the highlighted genes. Heatmap analysis further illustrated the direction and magnitude of transcript–protein changes across selected matched molecules (Fig. 5D). Spearman correlation analysis indicated predominantly positive mRNA–protein correlations, with correlation coefficients ranging from 0.600 to 0.943. Significant correlations were observed for Ndrg4 (Spearman r = 0.886, P < 0.05) and Gvin2 (Spearman r = 0.943, P < 0.05) (Fig. 5E).

Fig. 5.

Integrated transcriptome–proteome analysis prioritizes contractile, Ca²⁺ handling and metabolic pathways in DCM myocardium. A Venn diagram showing the overlap between quantified transcripts and proteins and differentially expressed mRNAs and proteins in DCM versus CT. B Distribution of gene-wise Spearman correlation coefficients between matched mRNA and protein abundances across samples, with negative and positive correlations shown. C Scatter plot of protein log2(DCM/CT) versus mRNA log2(DCM/CT) for matched gene–protein pairs. Points are colored by differential-status categories, including NDEPs_NDEGs, NDEPs_DEGs, DEPs_NDEGs, and DEPs_DEGs. The overall Spearman correlation coefficient is indicated. D Heatmap showing log2 fold changes of representative matched gene–protein pairs across the proteome and transcriptome layers. E Heatmaps of normalized protein and mRNA abundances for selected concordant molecules, together with their corresponding gene-wise Spearman correlations. F Quadrant plot illustrating concordant up- or down-regulation, oppositely regulated significant pairs, filtered pairs, and single-omics significant changes based on mRNA and protein log2 ratios. G Gene Ontology enrichment results across biological process, cellular component, and molecular function for the concordant set. H Gene–pathway linkage plot connecting representative concordant molecules to enriched pathways. I Pathway enrichment summary for the concordant set displayed as enrichment score, defined as −log10(p value)

Despite the limited overall concordance, the concordant subset demonstrated highly aligned regulation across layers. Hierarchical clustering revealed coordinated transcript–protein expression patterns (Fig. S2A), and their internal correlation within the concordant set was substantially stronger than the global relationship (Spearman R = 0.8947, Fig. S2B). Functionally, enrichment analysis of concordant molecules prioritized cardiomyocyte contractile machinery and Ca²⁺-linked programs, including sarcomere and myofibril components as well as calcium channel and transmembrane ion transport activities. In addition, several metabolic pathways were enriched, including branched-chain amino acid degradation, β-alanine metabolism, and pantothenate and CoA biosynthesis (Fig. 5G–I). In summary, these integrative analyses indicate broad cross-layer discordance in DCM myocardium, whereas a limited concordant module consistently points to remodeling of contractile machinery, Ca²⁺ handling, and metabolism, closely aligning with the in vivo cardiac dysfunction phenotype.

Metabolomic profiling reveals remodeling of choline–glycerophospholipid metabolism and central carbon pathways in DCM myocardium

Untargeted metabolomic profiling demonstrated a clear separation between CT and DCM myocardium in PCA space, indicating robust global metabolic remodeling in DCM hearts (Fig. 6A). Based on predefined statistical thresholds (fold change ≥ 1.5 or ≤ 0.67, p < 0.05), differential analysis identified 65 significantly altered metabolites in negative ion mode (35 increased and 30 decreased) and 60 in positive ion mode (21 increased and 39 decreased) in DCM versus CT (Fig. 6B, E–G). Pathway enrichment analysis demonstrated prominent perturbations in choline and glycerophospholipid metabolism. Retrograde endocannabinoid signaling, pathways related to insulin secretion and insulin resistance, and central carbon metabolism, including the pentose phosphate pathway (PPP), were also significantly altered (Fig. 6C). A metabolite–pathway interaction network further illustrated that these pathways were interconnected through shared metabolite hubs, suggesting coordinated rewiring of membrane lipid remodeling and central carbon flux (Fig. 6D).

Fig. 6.

Untargeted metabolomics identifies distinct metabolic remodeling in DCM myocardium with enrichment of choline–glycerophospholipid and central carbon pathways. A Principal component analysis score plot showing separation of CT and DCM myocardial metabolomes, with PC1 explaining 27.2% and PC2 explaining 16.7% of the variance. B Differential metabolites between groups displayed as log2 fold change. Metabolites meeting the predefined criteria of |FC| ≥ 1.5 and p < 0.05 are shown and colored by direction of change. C KEGG pathway enrichment analysis of differential metabolites. Bar length denotes −log10(p value), and color indicates the rich factor. D Metabolite–pathway network linking representative differential metabolites to enriched KEGG pathways. Node size reflects connectivity. E Heatmaps of representative choline- and glycerophospholipid-related metabolites across individual samples, shown as row-scaled z-scores. F Heatmap of representative amino acid and central carbon–related metabolites. G Heatmap of representative pentose phosphate pathway–related metabolites. Sample grouping is indicated above the heatmaps. Colors represent relative abundance expressed as z-scores.

At the metabolite level, DCM myocardium exhibited marked alterations in choline-containing metabolites and glycerophospholipid-related species, as evidenced by changes in glycerophosphocholine, phosphocholine, and multiple phosphatidylcholine (PC) and lysophosphatidylcholine (LPC) species (Fig. 6E). Concurrently, metabolites associated with amino acid handling and central carbon metabolism, including lysine, histidine, and pyruvate, showed consistent shifts (Fig. 6F). Metabolites linked to the PPP, such as 2-deoxyribose-5-phosphate, were also differentially abundant between groups (Fig. 6G). Collectively, these metabolomic findings indicate that DCM hearts undergo coordinated metabolic remodeling centered on choline–glycerophospholipid pathways and central carbon metabolism and PPP programs, providing a metabolic framework that aligns with systemic glucose intolerance, insulin resistance, and inflammatory activation observed in vivo.

Tri-omics integration reveals coordinated lipid remodeling, inflammatory and adhesion programs, and central carbon reprogramming in DCM myocardium

To obtain a systems-level view of diabetic cardiomyopathy remodeling, we integrated myocardial metabolomic, transcriptomic, and proteomic datasets and evaluated cross-layer convergence at both pathway and feature levels. Cross-omics KEGG pathway comparison identified membrane lipid remodeling as a predominant shared signature, with glycerophospholipid metabolism consistently dysregulated across all three molecular layers (Fig. 7A). Pathways related to systemic glucose dysregulation, including insulin secretion and insulin resistance, as well as carbon reallocation for redox support, such as the pentose phosphate pathway and central carbon metabolism, were also recurrently altered, aligning with in vivo observations of impaired glucose homeostasis and heightened inflammatory activation.

Fig. 7.

Integrated metabolomic, transcriptomic and proteomic analyses reveal convergent pathway remodeling and cross-layer correlation networks in DCM myocardium. A Cross-omics comparison of KEGG pathway enrichment in DCM versus CT showing pathways identified from proteomics, transcriptomics, and metabolomics. Bars are color-coded by omics layer, with proteomics in blue, transcriptomics in red, and metabolomics in green. Bar length reflects enrichment significance expressed as −log10(p value). B Correlation heatmap of selected genes, proteins, and metabolites across the three omics layers in relation to metabolite chemical classes (Class I). Colors denote correlation coefficients, with red indicating positive and blue indicating negative correlations. Hierarchical clustering highlights shared co-variation patterns across layers. C Integrated correlation network linking representative genes, proteins, and metabolites. Nodes denote feature type, edges indicate positive or negative correlations as color-coded, and node size reflects connectivity.

Notably, several biological signals exhibited layer preference: retinol and xenobiotic or drug metabolism by cytochrome P450 were more prominent at the proteome level, retrograde endocannabinoid signaling was primarily driven by metabolite changes, and circadian entrainment was most evident at the transcriptome level (Fig. 7A). In addition, immune and tissue remodeling pathways, including complement and coagulation cascades and adherens junction-related programs, were enriched predominantly in the transcriptome and proteome, consistent with immune and adhesion signatures identified in individual omics analyses (Fig. 7A).

Feature-class correlation profiling further demonstrated that metabolite classes segregated into distinct co-variation blocks with gene and protein signatures. Lipid classes, including glycerophospholipids, glycerolipids, fatty acyls, and sphingolipids, generally showed positive covariation with a broad range of transcripts and proteins, whereas several small-molecule classes, such as nucleotides and carboxylic acids, exhibited inverse correlation patterns (Fig. 7B). The integrated correlation network highlighted hub-like relationships linking choline-containing lipid metabolites, including glycerophosphocholine, phosphocholine, and multiple phosphatidylcholine and lysophosphatidylcholine species, with inflammation-associated gene and protein features, while contractile proteins formed an opposing module (Fig. 7C). Central carbon intermediates, such as pyruvate and metabolites related to the pentose phosphate pathway, including 2-deoxyribose-5-phosphate, served as bridge nodes connecting carbon flux pathways with broader lipid and immune programs (Fig. 7C).

To further contextualize these coordinated changes, integrated KEGG pathway maps were generated using Pathview. Glycolysis and pyruvate metabolism showed concordant reductions across multiple enzymatic nodes, whereas the pentose phosphate pathway displayed selective alterations consistent with carbon redistribution toward redox-supporting metabolism (Fig. S3A–B). Mapping onto the glucagon signaling pathway further revealed broad remodeling of upstream signaling and downstream metabolic control nodes (Fig. S3C), providing mechanistic context for the insulin secretion and insulin resistance signatures identified in cross-omics enrichment analyses (Fig. 7A).

Collectively, tri-omics integration delineates a coherent remodeling landscape in DCM myocardium characterized by membrane lipid and choline pathway remodeling, amplification of immune and adhesion programs, and reprogramming of central carbon metabolism.

Targeted validation confirms cholinergic remodeling, altered pyruvate handling and cGAS-STING activation in DCM myocardium

To validate key nodes emerging from the multi-omics analyses, we performed targeted biochemical, transcriptional, and histological assays in CT and DCM myocardium. DCM hearts displayed increased acetylcholine (ACh) content, accompanied by reduced acetylcholinesterase (AChE) mRNA levels and marked induction of choline O-acetyltransferase (ChAT) and the high-affinity choline transporter 1 (SLC5A7), providing evidence for remodeling of the myocardial cholinergic axis (Fig. 8A–D). In parallel, myocardial pyruvate content was significantly elevated, together with increased expression of pyruvate dehydrogenase kinase 4 (PDK4), indicating altered pyruvate handling and central carbon remodeling in DCM hearts (Fig. 8E–F). Consistent with contractile remodeling, myosin heavy chain 7 (MYH7) expression was reduced, whereas caveolin-3 (CAV3) remained unchanged (Fig. 8G–H).

Fig. 8.

Targeted validation confirms cholinergic remodeling, altered pyruvate handling, and activation of the cGAS–STING axis with increased collagen deposition in DCM myocardium. A Myocardial acetylcholine content normalized to total protein. (B–D) RT-qPCR quantification of acetylcholinesterase (AChE) B, choline acetyltransferase (ChAT) C and high-affinity choline transporter 1 (SLC5A7) D mRNA levels in the left ventricle from CT and DCM mice. E Myocardial pyruvate content normalized to total protein. (F–H) RT-qPCR quantification of pyruvate dehydrogenase kinase 4 (PDK4) F, myosin heavy chain 7 (MYH7) G and caveolin-3 (CAV3) (H) mRNA levels. I Representative immunofluorescence staining of cGAS (red), STING (green), and collagen type I (Col1a, red) with DAPI (blue) in left ventricular sections from CT and DCM mice. Scale bar indicates 20 μm. (J–L) Quantification of mean fluorescence intensity for cGAS J, STING K, and Col1a L shown as fold change. Data are presented as mean ± SD. Each dot represents one mouse. Statistical analysis was performed using an unpaired two-tailed Student’s t-test. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. ns indicates not significant

At the tissue level, immunofluorescence staining revealed increased cGAS and STING signals in DCM myocardium, along with enhanced collagen type I (Col1a) staining, consistent with activation of nucleic acid–sensing inflammatory signaling and increased collagen deposition (Fig. 8I–L). Collectively, these validation experiments corroborate the multi-omics–identified signatures of cholinergic and central carbon remodeling, cGAS–STING pathway activation, and fibrotic remodeling in DCM.

Discussion

In this study, we systematically characterized myocardial remodeling in an HFD + STZ-induced DCM mouse model. Although HFD/STZ-induced diabetic cardiomyopathy is frequently characterized by diastolic dysfunction with preserved LVEF, particularly during earlier disease stages, the cardiac phenotype varies among experimental protocols. Differences in dietary duration and composition, STZ administration, animal age, metabolic severity, and disease duration may contribute to this variability. Indeed, reductions in EF and/or fractional shortening have also been reported in HFD/STZ-induced diabetic models, suggesting that systolic impairment can occur under certain experimental conditions. Therefore, the reduced LVEF observed in the present study likely reflects a degree of systolic involvement in our experimental cohort rather than a universal feature of the HFD/STZ model. By integrating transcriptomic, proteomic and metabolomic datasets, we found that the diabetic myocardium undergoes coordinated molecular reprogramming along a mitochondrial–metabolism–inflammation–fibrosis axis. Inflammatory, immune and adhesion pathways were broadly activated, whereas glycolytic and pyruvate-related flux, calcium handling and excitation-contraction coupling were consistently suppressed. Membrane phospholipid–choline remodeling and enhancement of the pentose phosphate pathway appear to function as an integrative hub connecting energy metabolism, redox balance and structural remodeling. Targeted biochemical and molecular assays validated these multi-omics signatures and underpin a working model wherein chronic metabolic stress drives a shift toward an inflammation-dominant, electrically and mechanically vulnerable state in diabetic cardiomyopathy.

In this model, diabetic mice showed impaired systolic and diastolic function with left ventricular dilation, unchanged right ventricular free-wall thickness, and a higher heart weight-to-tibia length ratio, together with interstitial and perivascular fibrosis. Although DCM mice exhibited increased body weight, likely reflecting the effects of long-term high-fat diet feeding, tibia length was comparable between groups. Importantly, the heart weight-to-tibia length ratio remained significantly elevated in DCM mice, supporting cardiac hypertrophic remodeling independent of differences in skeletal body size. These morphometric findings (Fig. S4) were consistent with the structural and functional abnormalities identified by echocardiography and histological analyses. Ultrastructural analysis revealed mitochondrial swelling with disrupted cristae (Fig. 1), together indicating adverse remodeling of both the contractile apparatus and the energetic infrastructure. These cardiac changes occurred on a background of systemic insulin resistance, elevated circulating inflammatory cytokines, and increased myocardial inflammatory cytokine levels (Fig. 2), aligning with a state of chronic metabolic and inflammatory stress [23, 24]. Similar patterns of ventricular dysfunction, fibrosis, mitochondrial damage, and insulin resistance have been reported in other experimental models of diabetic cardiomyopathy and in patients with type 2 diabetes [23–26], supporting the translational relevance of our model. The high-fat diet plus single-dose streptozotocin protocol more closely mirrors the human pathophysiology of concurrent obesity, insulin resistance, and progressive β-cell dysfunction than do genetic models, for example db/db, or severe toxin-only models [25–27]. Thus, this integrated phenotypic characterization provides an essential anchor for interpreting the multi-omics signatures observed in the diabetic myocardium.

At the molecular level, our integrated transcriptomic and proteomic analysis points to a selective yet robust suppression of a contraction–calcium handling–mitochondrial energetics module in diabetic hearts. Transcriptomic profiling showed broad downregulation of gene sets related to sarcomeric structure, excitation–contraction coupling, calcium signalling, and oxidative phosphorylation. This was accompanied by reduced expression of key glycolytic and pyruvate metabolism genes, suggesting diminished delivery of glycolytic carbon into mitochondrial oxidative metabolism (Fig. 3). Data-independent acquisition-based quantitative proteomics corroborated these findings, demonstrating decreased abundance of several myosin light chains (e.g., Myl1, Myl4, Myl7), additional sarcomeric and Z-disk proteins, selected calcium-handling proteins, and multiple components of the tricarboxylic acid cycle and respiratory chain (Fig. 4). Multi-omics integration further indicated that, although the global correlation between mRNA and protein levels was modest, transcripts and proteins within this contraction–calcium–mitochondrial module tended to change in the same direction (Fig. 5), reinforcing the notion of coordinated downregulation across this functional axis. These findings are in line with clinical and experimental reports of impaired excitation–contraction coupling, depletion of myofilament proteins, and defective mitochondrial oxidative phosphorylation in diabetic cardiomyopathy and heart failure [23, 24, 28, 29]. The limited overlap between differentially expressed transcripts and proteins may reflect post-transcriptional regulation, differences in analytical sensitivity, myocardial heterogeneity, and limited statistical power. Therefore, cross-omics concordance was interpreted mainly at the pathway level. Among the concordant candidates, NDRG4 was decreased and GVIN2 was increased in DCM myocardium, and both were independently validated at the RNA and protein levels (Fig. S7). NDRG4 may be related to cardiac stress responses, whereas GVIN2 may reflect enhanced immune-inflammatory activity. Their precise roles in diabetic cardiomyopathy require further investigation.Within the context of our broader multi-omics analysis, this depressed contraction–calcium–mitochondrial module constitutes one pole of a remodelling spectrum that contrasts with activation of immune and adhesion pathways. It thereby provides a plausible mechanistic link connecting bioenergetic insufficiency to the observed structural and inflammatory alterations in the diabetic myocardium.

Our metabolomic profiling further supports a redistribution of central carbon flux from glycolysis and pyruvate oxidation toward the pentose phosphate pathway in diabetic hearts. Multiple glycolytic and TCA intermediates were altered, and metabolites related to pyruvate entry into mitochondrial oxidation were reduced, consistent with a bottleneck at the glycolysis–pyruvate–TCA interface (Fig. 6). Conversely, several pentose phosphate pathway intermediates were increased, together with metabolites related to nucleotide synthesis and glutathione metabolism. This pattern indicates that a portion of glucose-derived carbon is diverted into the pentose phosphate pathway to generate NADPH and biosynthetic precursors [30, 31]. Integrated pathway enrichment analysis across the transcriptome, proteome, and metabolome repeatedly highlighted the pentose phosphate pathway, nucleotide metabolism, and antioxidant pathways (Fig. 7), supporting a coordinated shift in central carbon metabolism rather than isolated changes. This metabolic signature aligns with the concept that under chronic nutrient excess, the diabetic heart enhances pentose phosphate pathway activity to maintain redox homeostasis and anabolic demands at the expense of oxidative ATP production [30–32]. Similar rebalancing of glucose utilization between glycolysis, the pentose phosphate pathway, and mitochondrial oxidation has been reported in diabetes or heart failure and in other contexts of cardiac remodeling, including pregnancy-associated and chemotherapy-related cardiac remodeling [32, 33]. A key novelty of our work is that, in an HFD + STZ-induced DCM mouse model, we directly demonstrate at the tri-omics level that the pentose phosphate pathway and related nucleotide and glutathione pathways are upregulated in parallel with suppression of pyruvate oxidation and TCA cycle activity. This multi-layer signature reflects a chronic metabolic stress phenotype in which central carbon reallocation is tightly coupled to energetic insufficiency and a pro-inflammatory, pro-remodeling myocardial milieu.

Our integrated multi-omics data further highlight membrane phospholipid and choline metabolism as a central hub linking metabolic remodeling with inflammation and structural alterations in diabetic hearts. Metabolomic profiling showed marked alterations in multiple glycerophosphocholine, lysophosphatidylcholine, and phosphatidylcholine species, indicating extensive remodeling of choline-containing glycerophospholipids (Fig. 6E). Consistently, the transcriptomic and proteomic datasets revealed coordinated regulation of enzymes involved in phospholipid synthesis, remodeling, and hydrolysis, together with transporters and lipid-handling proteins (Figs. 3, 4 and 5). Multi-omics network integration further demonstrated that these phospholipid and choline metabolites were densely connected with genes and proteins related to innate immune and inflammatory signaling, complement and coagulation cascades, cell adhesion molecules, and calcium-handling and contractile proteins (Fig. 7C). These findings support the concept that cardiac phospholipids function not only as structural components of cell membranes but also as dynamic reservoirs for bioactive lipid mediators and as platforms that organize calcium handling and signal transduction. Consequently, their remodeling can influence both electrical and mechanical stability and immune and fibrotic responses [34]. Recent experimental and clinical studies have implicated lysophosphatidylcholine and related choline-containing phospholipids in myocardial injury, inflammation, and adverse remodeling in diabetic cardiomyopathy and heart failure [35, 36]. Our results extend this work by demonstrating that, in the HFD + STZ-induced DCM mouse model, phospholipid and choline remodeling sits at the intersection of central carbon metabolism and immune and structural pathways, emerging as a plausible mechanistic link between metabolic stress, inflammatory activation, and ventricular dysfunction in diabetic cardiomyopathy.

Across the three omics layers, we observed broad amplification of immune and inflammatory programs in DCM hearts. Enriched pathways included cytokine–cytokine receptor interaction, chemokine signaling, NOD-like receptor signaling, complement and coagulation cascades, and cell adhesion molecules (CAMs). Transcriptomic analysis in particular showed that upregulated genes were enriched for pathogen-response and interferon signaling pathways, together with cytosolic DNA-sensing and phagosome-related programs. This was accompanied by increased expression of adhesion molecules and activation of NF-κB-related signaling, indicating a molecular basis for inflammasome activation, interferon signaling, and enhanced immune cell adhesion and recruitment in the diabetic myocardium [37, 38]. In our model, this omics signature was paralleled by systemic insulin resistance, elevated circulating inflammatory mediators, and increased myocardial expression of proinflammatory markers. Together, these observations support a proinflammatory and prothrombotic cardiac microenvironment that favors fibrotic remodeling [37, 38]. Complement-related pathways, which bridge innate immunity and metabolic stress, have also been implicated in cardiovascular and metabolic diseases including DCM and may contribute to coupling between adipose-derived signals and myocardial injury [39]. The prominent innate immune and DNA-sensing signatures identified by our multi-omics analyses prompted us to examine the cGAS–STING axis as a representative component of the inflammatory remodeling observed in DCM. Consistent with previous studies, we confirmed increased cGAS–STING signaling in DCM myocardium, accompanied by marked mitochondrial ultrastructural abnormalities and enhanced oxidative stress. Importantly, activation of cGAS–STING itself is not unique to the present study. Previous mechanistic work has demonstrated that mitochondrial injury and the release of mitochondrial DNA into the cytosol can activate cGAS–STING signaling in diabetic hearts and promote downstream inflammatory responses, including NLRP3 inflammasome activation, pyroptosis, hypertrophic remodeling, and cardiac dysfunction; conversely, genetic or pharmacological inhibition of this pathway can attenuate DCM progression [40, 41]. In this context, our findings extend these observations by positioning cGAS–STING activation within a broader multi-omics landscape characterized by coordinated mitochondrial dysfunction, metabolic reprogramming, and immune-inflammatory remodeling. Thus, rather than establishing a new causal mechanism for cGAS–STING in DCM, our data provide complementary molecular evidence that this pathway is integrated into the broader mitochondria–metabolism–inflammation network identified in the HFD/STZ-induced DCM model. Although direct quantification of mitochondrial DNA release was beyond the scope of this study, the convergence of multi-omics evidence for DNA-sensing pathway activation, histological evidence of mitochondrial damage, and prior mechanistic work supports a model in which mitochondrial stress engages cGAS–STING-mediated sterile inflammation. This inflammatory node then converges with complement, coagulation, and adhesion pathways to amplify immune cell recruitment, endothelial activation, and fibroblast-driven remodeling in the diabetic heart.

Synthesizing these findings, we propose a working model in which the diabetic myocardium undergoes remodeling along a mitochondrial–metabolism– inflammation–fibrosis axis [37, 42]. One branch is characterized by extensive remodeling of membrane phospholipid–choline metabolism and a shift in central carbon flux from pyruvate oxidation toward the pentose phosphate pathway, which supports NADPH generation and biosynthesis. Under chronic nutrient excess, this reallocation may also predispose to excess reactive oxygen species and secondary mitochondrial DNA stress [41, 43, 44]. These mitochondrial danger signals can activate the cGAS–STING pathway and related innate immune sensors, amplifying cytokine production, complement and coagulation activation, immune recruitment programs, and fibroblast remodeling, thereby perpetuating a chronic proinflammatory and profibrotic milieu in the diabetic heart [37, 42, 43]. On the other branch, our multi-omics data indicate coordinated suppression of calcium handling, sarcomeric structure, and antioxidant and detoxification pathways, together with impaired mitochondrial oxidative metabolism. This aligns with established concepts of metabolic dysregulation in diabetic cardiomyopathy and further undermines electromechanical stability, as well as the energetic and antioxidant reserve required to withstand sustained stress [41, 44]. We postulate that these two branches engage in a vicious cycle. Metabolic and mitochondrial dysfunction fuels innate immune activation and adverse structural remodeling, while persistent inflammation, endothelial dysfunction, and fibrosis further impair excitation–contraction coupling and tissue oxygenation. This feed-forward interaction may drive the transition from a compensated state to an irreversible, inflammation-dominant, and energetically fragile state characteristic of advanced diabetic cardiomyopathy.

Our integrated analysis yields several translational implications. The integrated multi-omics map highlights phospholipid–choline metabolism, the pentose phosphate pathway and central carbon metabolism, the cGAS–STING innate immune axis, and the calcium-handling–mitochondrial energetics module as promising circuits in diabetic cardiomyopathy. These pathways intersect with mechanisms increasingly recognized as therapeutic targets in diabetes and heart failure, including modulation of myocardial substrate use, mitochondrial function, and inflammation by sodium–glucose cotransporter 2 inhibitors and, more recently, glucagon-like peptide-1 receptor agonists [45, 46]. Emerging single-cell transcriptomic studies in heart failure suggest that empagliflozin may modulate cardiomyocyte communication, highlighting intercellular signaling as another layer of cardiac remodeling [47]. In this context, phospholipid- and choline-related species, such as lysophosphatidylcholines, selected pentose phosphate pathway intermediates, and circulating markers of mitochondrial injury or cGAS–STING activation hold promise as biomarkers to refine risk stratification and to monitor responses to metabolic or immunomodulatory interventions in patients with DCM [35, 41, 48]. Consequently, pharmacologic strategies that specifically target phospholipid–choline remodeling, redirect central carbon flux, or attenuate maladaptive cGAS–STING signaling could complement neurohormonal blockade and sodium–glucose cotransporter 2 inhibition, and warrant further evaluation in preclinical models.

Several limitations of this study must be acknowledged. First, our analysis was conducted in a single rodent species, primarily in one sex, and at a single late disease time point. It also did not resolve cell-type-specific contributions. These design choices may limit our ability to capture the full cellular and temporal heterogeneity of the diabetic heart. Second, the multi-omics analyses are largely correlative and rely on pathway and network enrichment. Therefore, causal links between the identified molecular changes and cardiac dysfunction remain to be demonstrated. Third, although we provide evidence for mitochondrial injury and activation of cGAS–STING signaling, we did not directly quantify mitochondrial DNA release or dissect cell-type-specific sources of cytosolic DNA. More detailed mechanistic studies with genetic and pharmacologic modulation of this pathway are needed. Future work should incorporate temporal and single-cell-resolved profiling, targeted interventions focusing on the pentose phosphate pathway, phospholipid–choline remodeling, and cGAS–STING signaling, as well as validation of key signatures in human myocardial tissue and plasma. Such efforts will be essential to translate the present multi-omics framework into concrete biomarkers and therapeutic strategies for patients with diabetic cardiomyopathy. Finally, the transcriptomic and proteomic analyses were based on only three biological replicates per group. Although PCA and sample-correlation (Fig. S6) confirmed data consistency, the small sample size may reduce statistical power for modest effects and limit generalizability. Therefore, these results are best interpreted at the pathway and cross-omics level, and individual candidates require validation in larger cohorts.

Some individual molecular changes were not fully concordant with the observed phenotypes. Although Nppa and SERPINE1 were downregulated, several remodeling-related genes, including Bmp4, Ccr2, Cybb, and Mmp3, were upregulated, supporting the overall hypertrophic, inflammatory, and fibrotic remodeling phenotype. In addition, the three omics analyses were performed on randomly selected subsets of the same experimental cohort rather than identical animals, which may contribute to inter-individual variability. Therefore, cross-omics concordance was interpreted primarily at the pathway and functional-network levels rather than by strict one-to-one agreement among individual molecules.

In summary, this study integrates myocardial transcriptomic, proteomic, and metabolomic profiles with detailed phenotyping to describe how long-standing type 2 diabetes remodels the heart. We show that DCM is characterized by coordinated alterations along a mitochondria–metabolism–inflammation–fibrosis axis. Within this axis, we identify pervasive remodeling of membrane phospholipid–choline metabolism, a shift in central carbon flux favoring the pentose phosphate pathway, activation of the cGAS–STING innate immune pathway, and suppression of the calcium-handling–mitochondrial energetics module as key interconnected components. Within this framework, membrane phospholipid–choline remodeling and diversion of central carbon flux toward the pentose phosphate pathway emerge as metabolic hubs that link mitochondrial stress to innate immune activation and contractile dysfunction. These insights provide a structured basis for future mechanistic studies and for the development of biomarker and therapeutic strategies targeting specific nodes within this network in diabetic cardiomyopathy.

Authors' contributions

Feng Zhang performed the majority of experiments, coordinated sample collection, analyzed and interpreted the data, and drafted the manuscript. Chunbo Ai contributed to key experiments, data interpretation, and manuscript revision. Hao Zhang and Xiaoqiang Sun assisted with omics-related experiments, animal studies, and phenotyping, and contributed to data acquisition. Li Wang, Chao Li, and Wei Cai supported data processing and statistical and bioinformatic analyses, assisted with cross-omics integration, and provided technical support for proteomic and metabolomic workflows. Xuemei Yin, Yanxin Wang, and Yanwei Li contributed to histology and imaging analyses, model-related experiments, and data collection. Chengzhi Lu and Zhuqing Li conceived and supervised the study, interpreted results, and critically revised the manuscript. All authors read and approved the final version of the manuscript.

Funding

This research was funded by the National Natural Science Foundation of China (82300445), the Medical Health Science and Technology Project of Tianjin Municipal Health Commission (TJWJ2023QN026), and Research Plan Project of Hebei Provincial Administration of Traditional Chinese Medicine (T2025011).

Data availability

The RNA-sequencing data generated in this study are publicly available in the NCBI Sequence Read Archive (SRA) under BioProject accession number PRJNA1494032 (https://www.ncbi.nlm.nih.gov/bioproject/PRJNA1494032). The proteomics data are publicly available through the ProteomeXchange Consortium via the iProX partner repository under dataset identifier PXD081020 and iProX project accession IPX0018335000 (https://proteomecentral.proteomexchange.org/cgi/GetDataset? ID=PXD081020). The metabolomics data are publicly available in the MetaboLights repository under study accession number MTBLS15043 (https://www.ebi.ac.uk/metabolights/editor/MTBLS15043/overview). All other data supporting the findings of this study are included in the article and its Supplementary Information.

Declarations

Ethics approval and consent to participate

All animal procedures were conducted in accordance with the guidelines for the care and use of laboratory animals and were approved by the Institutional Animal Care and Use Committee (IACUC) of Nankai University. All efforts were made to minimize animal suffering and to use the minimum number of animals necessary to obtain reliable results.

Consent for publication

Not applicable.

Competing interests

The authors declare no conflict of interest.