The discovery cohort included 38 bulk RNA-sequencing samples from male C57BL/6J mice divided among four groups: controls on standard chow, high-fat-diet controls, diabetic mice and diabetic mice treated with insulin glargine. Diabetes was induced using six to eight weeks of a high-fat diet containing 60% of calories from fat, followed by five daily injections of streptozotocin. A mouse was classified as diabetic when random blood glucose exceeded 16.7 mmol/L, or 300 mg/dL, on three consecutive measurements. Control and high-fat-diet groups were sampled at all six biweekly intervals, diabetic mice were represented at four intervals after model establishment and RNA quality control, and insulin-treated mice at three intervals because treatment began after the first sampling.

Principal component analysis, used descriptively, separated control from diabetic heart samples as the diabetic group shifted over time, while high-fat-diet samples occupied an intermediate position and insulin-treated samples fell between the control and diabetic clusters. The separation was more pronounced in the kidney, where diabetic samples showed a greater distance from controls. Differential expression analysis, using thresholds of more than a twofold change and an adjusted P value below 0.05, likewise produced more changes in kidney than in heart across the major contrasts. The diabetes-versus-control comparison contained the largest set of kidney-specific changes, and gene ontology enrichment linked those changes to antigen processing and presentation, inflammatory response and leukocyte activation, with false discovery rates below 0.05.

Regression-based maSigPro analysis was then used to identify genes with significant time-dependent profiles. Its thresholds were a Q value below 0.3 and an R-squared value above 0.6, applied as an exploratory screen because the discovery RNA-sequencing cohort included only one animal per group per time point. The kidney yielded 74 unique core time-dependent genes retained across the major renal contrasts. Hierarchical clustering divided them into four groups. The largest contained immune and antigen-presentation genes, including Cxcl10, Icam1, B2m, H2-K1 and H2-D1, while another included metabolic genes such as Mogat2 and Enpp2. Trajectory plots indicated that several inflammatory transcripts rose early in diabetes and later declined. Insulin altered these trajectories during selected intervals, while protein-interaction analysis placed H2-D1, H2-K1, B2m and Cxcl10 at the center of the principal kidney network.

The cardiac response was smaller and followed a different functional pattern. Twenty-two time-dependent cardiac genes were enriched in both immune pathways and metabolic processes. Significant immune terms included leukocyte-mediated cytotoxicity and T-cell-mediated cytotoxicity, with Prf1, Gzmb, Fas and Cd8a among the contributing genes. Metabolic pathways included fatty-acid derivative biosynthesis, while Hmgcs2, Gcdh and Scd1 represented ketogenesis, fatty-acid oxidation and lipid desaturation. The contrast between organs was therefore not simply one of overall disease severity: kidney trajectories were dominated by immunological remodeling, whereas the heart combined immune activation with prominent metabolic adaptation. The six-visit window covered roughly 10 weeks of early-to-intermediate disease in the high-fat-diet and streptozotocin model, not the full course of human cardiorenal disease.

To examine cellular heterogeneity, the researchers analyzed a separate public single-cell RNA-sequencing dataset comprising 16 mouse kidneys from control, diabetic kidney disease and SGLT2 inhibitor treatment groups. After quality control, 18 renal cell types were resolved, including proximal convoluted and straight tubules, injured proximal tubule cells, podocytes, endothelial cells, fibroblasts and immune cells. Enpp2 expression was significantly higher in injured proximal tubule cells from diabetic kidneys than from controls, with a reported P value below 0.001. SGLT2 inhibitor treatment was associated with lower Enpp2 expression in that population. In contrast, the MHC class I transcripts H2-K1 and H2-D1 were lower in immune cells from diabetic kidneys, although comparisons between diabetic and treated kidneys were not uniformly significant for these two genes.

Inferred cell-cell communication networks suggested stronger signaling involving injured proximal tubule cells in diabetic kidneys and reduced interactions associated with fibroblasts, with some of these patterns shifted in the SGLT2 inhibitor group. An overlap analysis also connected the 74 time-dependent bulk kidney genes with cell-type-specific differentially expressed genes; H2-K1 and Plin2 were altered in both datasets. The parallel observations involving two different treatments are hypothesis-generating because insulin and SGLT2 inhibition came from separate cohorts with different designs. The single-cell analysis therefore helped resolve where selected transcripts were expressed, but it could not establish that the drugs had equivalent effects or support a direct ranking of therapeutic efficacy.

Independent quantitative PCR used three mice per group at time points two, three and four. In the heart, Mogat2 and Hmgcs2 showed time-dependent variation: Mogat2 increased in high-fat-diet and diabetic mice at the second time point, while insulin partially attenuated the initial increase and was followed by a later rise and a sharp drop at the fourth point. Hmgcs2 was elevated in diabetic mice from the second time point and followed a different pattern under insulin. Kidney immune markers Cxcl10, H2-D1 and H2-K1 increased significantly in diabetic mice at the second and third time points, with P values below 0.0001, before declining at the fourth. Insulin suppressed these markers during selected intervals. By contrast, Enpp2 remained stable early and rose significantly in diabetic mice at the fourth time point, while Icam1 and Plin2 stayed elevated across the assessed periods.

Protein-level measurements broadly supported selected RNA findings. Immunohistochemical H-scores for ENPP2, ICAM1 and PLIN2 were quantified across 15 randomly selected fields per group by observers blinded to treatment identity. ENPP2 was mainly localized to injured proximal tubular epithelial cells in diabetic kidneys, and insulin markedly reduced this signal. PLIN2 and ICAM1 staining increased in diabetic kidneys, and insulin significantly attenuated both elevations. Histological examination with hematoxylin and eosin and periodic acid-Schiff staining found glomerular and tubular alterations in diabetic kidneys, whereas insulin-treated kidneys more closely resembled controls. The results support a sequence in which early immune activation is followed by later lipid- and adhesion-related remodeling, with systemic glycemic control modifying parts of that sequence. The study does not establish that ENPP2 or PLIN2 directly causes injury, and it was not designed to test treatment benefit in humans.

Several limitations constrain how far the findings can be generalized. The time-series RNA-sequencing phase had one animal per group per time point, making it a discovery screen rather than a definitive population-level trajectory study, and only male mice were included. Although qPCR and immunohistochemistry provided independent validation for selected kidney markers and cardiac transcripts, cardiac proteins and cardiac histology were not assessed in the same way. The sampling window also ended before advanced disease, and no paired functional measures of cardiac or renal performance were reported. Larger studies including both sexes, direct testing of ENPP2 and PLIN2, and functional outcomes would be needed to determine whether the observed modules are causal or clinically predictive. For now, the study offers a time-ordered working model in which the diabetic kidney changes earlier than the heart, early immune activity gives way to structural and lipid-related programs, and insulin is associated with partial rather than complete modulation of selected molecular trajectories.

Subject of Research: Longitudinal transcriptomic differences between diabetic kidney and heart tissue in mice

Article Title: Integrated time-series multi-omics reveals temporal reprogramming and inter-organ divergence in diabetic cardiorenal syndrome

Article References: Gao, S.-S., Wan, G., Wang, Q., An, J., Yang, Q., Liu, C., & Cai, Y. (2026). Integrated time-series multi-omics reveals temporal reprogramming and inter-organ divergence in diabetic cardiorenal syndrome. Frontiers in Endocrinology, 17, Article 1950040. https://doi.org/10.3389/fendo.2026.1950040

Image Credits: AI Generated

DOI: 10.3389/fendo.2026.1950040

Keywords: diabetic cardiorenal syndrome, time-series RNA sequencing, single-cell transcriptomics, kidney injury, heart remodeling, ENPP2, PLIN2, ICAM1, insulin treatment, mouse model, Integrated, time-series