Transcriptome analysis of type 2 diabetic cardiomyopathy model: discovering the role of the key gene Slc16a3

After a 12-week feeding period with a diet of 60% high fat, Leprdb mice exhibited notable elevations in body weight, blood glucose, and triglyceride levels compared to control mice (m Leprdb) maintained on a standard diet. This precisely recapitulated the metabolic hallmarks of obesity, hyperglycemia, and hyperlipidemia observed in T2DM (Fig. 1A). Comprehensive investigations further unveiled that both Leprdb and T2DM patients exhibited notably elevated blood lactate levels relative to the control group (Fig. 1B). Echocardiographic assessments revealed a reduced E/A ratio in Leprdb mice, serving as a distinct indicator of diastolic dysfunction (Fig. 1C). Furthermore, pathological staining techniques highlighted increased myocardial thickness and broadened fibrotic areas in Leprdb mice (Fig. 1D). Collectively, these findings corroborate the successful establishment of a mouse model for type 2 diabetic cardiomyopathy.

Fig. 1

A Body weight, blood glucose and triglycerides in m Leprdb and Leprdb mice. B Blood lactic acid in Leprdb vs. m Leprdb mice and T2DM vs. non-T2DM. C Pulsed‐wave Doppler showing diastolic function: peak velocity of the E wave and A wave, E/A ratio in Leprdb vs. m Leprdb mice. D HE and Masson staining of the heart in Leprdb vs. m Leprdb mice. E Volcano plot showing differentially expressed genes (DEGs) of Leprdb vs. m Leprdb mice. F GO and KEGG analysis. G Protein–protein interaction (PPI) hub networks of Slc16a3. H MCT4 immunohistochemical staining of heart tissue in Leprdb vs. m Leprdb mice. I Cardiac Mct1, 3, 4 mRNA expression in Leprdb vs. m Leprdb mice. J Western blot analysis and quantification of MCT4, MCT1, and MCT4:MCT1 ratio in Leprdb vs. m Leprdb mice

We conducted transcriptome sequencing analysis on heart tissue samples obtained from both Leprdb and m Leprdb mice. By applying a threshold of an absolute Log2FC value greater than 0.585 and an adjusted P-value less than 0.05, we identified a total of 536 upregulated genes and 424 downregulated genes (Fig. 1E). Subsequent GSEA analysis revealed that these differentially expressed genes (DEGs) were predominantly enriched in biological processes related to interferon signaling, innate immunity, viral myocarditis, and extracellular matrix (ECM) regulation (Additional file 1: Fig. S1A). These findings suggest a potential association between DCM, innate immune responses, and inflammatory processes.

Concurrently, immune infiltration analysis revealed significantly higher proportions of monocytes and macrophages in Leprdb heart tissues than m Leprdb (Additional file 1: Fig. S1B), indicating a crucial role for monocyte-macrophages in the pathogenesis of DCM. Through GO analysis, we observed that these DEGs were primarily enriched in biological pathways involving fatty acid binding/oxidation, glycolysis, and monocarboxylic acid binding/transport. These processes are closely associated with oxidative stress response, cytokine activity, and chronic inflammation. Additionally, the KEGG analysis highlighted the involvement of these DEGs in biological processes such as carbon metabolism, HIF-1 signaling pathway, and dilated cardiomyopathy (Fig. 1F). These findings suggest a pivotal role for interconnected biological processes involving energy metabolism, oxidative stress, and inflammation in the development of DCM. Furthermore, by leveraging protein–protein interaction analysis and the cytohub plugin, we identified a key gene, Slc16a3. This gene exhibits associations with multiple genes implicated in glycolysis metabolism, including Ldha, Pfkl, Pfkp, and Tiger (Additional file 1: Fig. S1C, Fig. 1G). According to annotations from the UniProt database (https://www.uniprot.org/), the protein product of the Slc16a3 gene is MCT4, which possesses a pore structure (Fig. 1G) and facilitates the transport of monocarboxylic acids such as lactate.

Immunohistochemical experiments convincingly demonstrated a marked increase in MCT4 expression in Leprdb heart tissues compared to m Leprdb (Fig. 1H). This observation is consistent with our RT-qPCR analysis, which further substantiates the upregulated trend of Mct4 in the Leprdb (Fig. 1I). Notably, the RT-qPCR analysis revealed no significant variations in the expression of Mct1 and Mct3 between the two groups (Fig. 1I). Given that MCT4 and MCT1 are responsible for lactate efflux and influx, respectively, we further examined the expression of these two proteins in mouse heart tissue using Western blot and calculated their ratio. The results showed that in Leprdb mice, the expression level of MCT4 and the ratio of MCT4 to MCT1 were significantly higher than those in the control group. In contrast, the expression level of MCT1 remained stable between the two groups (Fig. 1J). These discoveries strongly implicate the upregulated expression of MCT4 as a potential key factor in the development of DCM.

We postulate that this phenomenon might be intimately linked to the process of lactate transport across the plasma membrane. To investigate this hypothesis, we further isolated the plasma membrane from heart tissue and re-evaluated MCT4 expression. The results were consistent, revealing a significantly higher expression of MCT4 in the heart plasma membrane of Leprdb mice compared to the control group (Additional file 1: Fig. S1D). These experimental observations collectively implicate a pivotal role of MCT4 located on the myocardial plasma membrane in regulating lactate transport and energy homeostasis, thereby hinting at its potential involvement in the pathophysiology of DCM.

Mechanism of fatty acid-induced cardiomyocyte injury: MCT4-mediated imbalance of lactate-pyruvate axis

Fatty acid overload and lipotoxicity are critical factors triggering myocardial injury in T2DM [14, 15]. To mimic an in vitro environment of hyperlipidemia, we utilized PA to stimulate primary mouse cardiomyocytes (PMCM) or H9C2 cells. Our findings revealed a time-dependent elevation of reactive oxygen species (ROS) levels in H9C2 cells following PA stimulation (Fig. 2A). Additionally, we observed a significant decrease in Λψm after 24 h of PA stimulation (Additional file 1: Fig. S2A, Fig. 2C). Similarly, elevated ROS levels were also observed in PMCM (Fig. 2B), indicating that free fatty acids are the primary drivers of mitochondrial oxidative stress in cardiomyocytes.

Fig. 2

A Representative diagram and quantification of ROS using DCFH-DA staining in H9C2 cells treated with palmitic acid (PA). B Representative diagram and quantification of ROS in PMCM cells treated with PA. C Quantification of Λψm (mitochondrial membrane potential) in H9C2 cells treated with PA. D Fold change of lactic acid in H9C2 whole cell lysates after time-dependent PA stimulation. E Fold change of lactic acid in H9C2 cell supernatant after time-dependent PA stimulation. F Mct4 mRNA expression in H9C2 cells treated with PA. G Western blot analysis and quantification of MCT4 in H9C2 cells treated with PA. H Fold change in whole-cell abundances of lactic acid in H9C2 treated with PA and/or VB124. I Fold change in cell supernatant abundances of lactic acid in H9C2 treated with PA and/or VB124. J Representative diagram and quantification of ROS in PMCM and H9C2 cells treated with PA and/or VB124. K Representative diagram and quantification of Λψm in H9C2 cells treated with PA and/or VB124. L Fold change of ATP content in H9C2 treated with PA and/or VB124

Previous research has highlighted the crucial role of the lactate-pyruvate axis in regulating cardiac hypertrophy and heart failure [16]. To explore the effects of PA stimulation on the dynamic changes of this axis in cardiomyocytes, we designed and conducted a series of time-gradient experiments in H9C2 cells. The findings uncovered an intriguing phenomenon: upon PA stimulation, the intracellular lactate level initially decreased but subsequently increased gradually. Concurrently, the pyruvate level demonstrated an opposite trend, first increasing and then declining (Fig. 2D, Additional file 1: Fig. S2B). Moreover, we observed that PA caused a time-dependent increase in lactate concentration in the supernatant of cardiomyocytes (Fig. 2E), suggesting that PA may induce an imbalance in the lactate-pyruvate axis and cellular lactate efflux in cardiomyocytes.

MCT4, a vital transporter for lactate in cardiomyocytes, is typically expressed at low levels under healthy conditions, as reported in previous research [17]. Nevertheless, our experimental findings indicate that both Mct4 mRNA and protein expression exhibited a time-dependent upregulation in H9C2 cells upon stimulation with PA (Fig. 2F, G). Additionally, we conducted a specific analysis of MCT4 expression in the plasma membrane fraction of H9C2 cells. The findings further confirm an enhanced expression of MCT4 on the plasma membrane in response to PA stimulation. (Additional file 1: Fig. S2C). These findings suggest that free fatty acids may disrupt the imbalance of the lactate-pyruvate axis in cardiomyocytes by inducing MCT4 upregulation. To validate this hypothesis further, we utilized VB124 to specifically inhibit MCT4 activity or employed siRNA to knock down MCT4 expression in H9C2 cells. The experimental results revealed that VB124 significantly reduced the PA-induced increase in lactate levels in whole cells and supernatants (Fig. 2H, I), decreased whole-cell pyruvate levels (Additional file 1: Fig. S2D), and increased mitochondrial pool pyruvate levels (Additional file 1: Fig. S2E). The effect of MCT4 knockdown was consistent with the treatment effect of VB124 (Additional file 1: Fig. S2F–I). Importantly, we also found that knockdown or inhibition of MCT4 could attenuate ROS production in cardiomyocytes induced by PA (Fig. 2J, Additional file 1: Fig. S2J), reverse the decrease of Λψm (Fig. 2K, Additional file 1: Fig. S2L), and enhance ATP production (Fig. 2L, Additional file 1: Fig. S2M). Furthermore, we detected mitochondrial superoxide using MitoSOX and similarly observed that inhibiting MCT4 reduced PA-induced mitochondrial superoxide production in both PMCM and H9C2 cardiomyocytes (Additional file 1: Fig. S2K). Taken together, these findings suggest that MCT4 upregulation may mediate the imbalance of the lactate-pyruvate axis induced by fatty acids, ultimately leading to mitochondrial oxidative stress damage in cardiomyocytes.

Protective effects of inhibiting/knocking down MCT4 on free fatty acid-induced cardiomyocyte injury

Given the intricate connections between oxidative stress injury, cardiac hypertrophy, inflammation, and apoptosis, we investigated the precise mechanisms of free fatty acid-induced damage in cardiomyocytes. We aimed to assess the potential protective effects of inhibiting or suppressing MCT4 against cardiomyocyte lipotoxicity. Utilizing immunofluorescence techniques, we quantified the expression levels of BNP in both PMCM and H9C2 cells. Notably, our results revealed a significant elevation of BNP expression upon PA exposure, which was effectively mitigated by VB124 treatment (Fig. 3A). Consistent with this observation, MCT4 knockdown in H9C2 cells exhibited a comparable effect to VB124 treatment (Additional file 1: Fig. S3A). Additionally, we examined the transcriptional profiles of Anp and Bnp, observing that both VB124 and MCT4 knockdown markedly reversed the PA-induced upregulation of Anp/Bnp mRNA (Fig. 3B, Additional file 1: Fig. S3B). By assessing the length-to-width ratio of cardiomyocytes, we further demonstrated that VB124 could substantially counteract the PA-induced reduction in this ratio in NMCM and H9C2 cells (Fig. 3C, Additional file 1: Fig. S3C), suggestive of an improvement in cardiac hypertrophy. Furthermore, we observed a marked increase in apoptosis rates in PA-exposed NMCM and H9C2 cells through apoptosis quantification, which was significantly attenuated by VB124 pretreatment (Fig. 3D, Additional file 1: Fig. S3D). Lastly, our analysis of inflammatory cytokine transcription levels revealed that pretreatment with either VB124 or MCT4 knockdown significantly reduced the mRNA levels of Il-1β, Il-6, Il-18, and Ccl2 in H9C2 cells challenged with PA (Fig. 3E, Additional file 1: Fig. S3E). Collectively, our findings reveal the protective role of MCT4 inhibition or knockdown in reversing PA-induced cardiomyocyte hypertrophy and apoptosis while concurrently mitigating the transcription of inflammatory cytokines.

Fig. 3

A Representative diagram and quantification of BNP in PMCM and H9C2 cells treated with PA and/or VB124. B Anp and Bnp mRNA expression in H9C2 cells treated with PA and/or VB124. C Representative diagram and quantification of length–width ratio in PMCM treated with PA and/or VB124. D Representative diagram and quantification of Annexin-V positive apoptosis cells in PMCM treated with PA and/or VB124. E mRNA expression of Il-1β, Il-6, Il-18, Ccl2 and Tnf-α in H9C2 cells treated with PA and/or VB124

Effects of MCT4-mediated lactate transport in cardiomyocytes on macrophage inflammatory response

After conducting a correlation analysis of RNA-seq expression profiles, a positive association was observed between the macrophage marker CD68 and MCT4 (Fig. 4A). Immunofluorescence staining further enabled us to visualize the colocalization of MCT4 and CD68. Notably, we identified a prominent accumulation of CD68+ macrophages in regions exhibiting elevated MCT4 expression within Leprdb heart tissue (Fig. 4B). These findings suggest that the upregulation of MCT4 in cardiomyocytes under T2DM conditions may play a role in promoting cardiac macrophage infiltration. To further elucidate how alterations in MCT4 within cardiomyocytes influence macrophages, we employed a Transwell co-culture method (Fig. 4C). Crystal violet staining indicated that when MCT4 in H9C2 cells was knocked down or inhibited, the migration of co-cultured macrophages induced by PA was significantly reduced (Fig. 4D, Additional file 1: Fig. S4A). Additionally, inhibiting MCT4 in NMCM cardiomyocytes also led to a substantial decrease in mRNA transcription levels of the inflammatory cytokines Il-1β and Tnf-α in PA-induced co-cultured macrophages (Additional file 1: Fig. S4B), as well as reduced mRNA transcription levels of the M1 macrophage polarization marker iNOS and hypoxia-inducible factor Hif-1α (Fig. 4E). Notably, the effects of MCT4 knockdown in H9C2 cells on macrophages were consistent with the impact of VB124 treatment (Additional file 1: Fig. S4C).

Fig. 4

A Spearman correlation analysis between MCT4 and CD68 mRNA expression in the heart of mice. B Representative diagram of CD68 expression around the high expression of MCT4 in the heart of m Leprdb and Leprdb mice. C A diagram to depict the coculture experiment using transwell. RAW264.7 and H9C2 (pre-treated with PA and/or VB124) were cocultured for 24 h for D–F. D Representative diagram and quantification of migrated RAW264.7 cells stained with crystal violet. E mRNA expression of iNOS and Hif-1α in coculture RAW264.7. F Representative diagram and quantification of Pan Kla% in coculture RAW264.7. G Flow cytometry analysis of RAW264.7 macrophage subtypes treated with either PA, lactic acid (LAC), or a combination of both. H Representative diagram and quantification of H3K18La and H4K12La in RAW264.7 treated with PA and/or LAC. I mRNA expression of Tnf-α and Il-1β in RAW264.7 treated with PA and/or LAC. J Hif-1α mRNA expression in RAW264.7 treated with PA and/or LAC. K Chromatin IP (ChIP) quantitative real time PCR (ChIP-qRT-PCR) of gene targets for HIF-1α and iNOS at H4K12La in RAW264.7 treated with PA or PA + LAC

Given the pivotal role of MCT4 in lactate transport and the fact that lactate serves as a critical precursor in the process of histone lactylation, we hypothesize that knocking down or inhibiting the activity of MCT4 can effectively reduce the release of lactate from cardiomyocytes, subsequently decreasing the level of histone lactylation in adjacent infiltrating macrophages. To test this hypothesis, we employed immunofluorescence analysis and found that inhibiting MCT4 in NMCM cardiomyocytes significantly attenuates the enhanced histone lactylation observed in co-cultured macrophages induced by PA (Fig. 4F), further supporting our speculation. Based on these findings, we further hypothesize that changes in histone lactylation in macrophages may be closely related to their polarization states. To explore this hypothesis, we used RAW264.7 macrophages and treated them with lactic acid (LAC), PA, or a combination of both. Subsequent flow cytometry analysis revealed that lactic acid alone increases the proportion of CD206+ anti-inflammatory macrophages while decreasing the proportion of CD86+ pro-inflammatory macrophages. Conversely, PA alone induces an increase in the proportion of CD86+ pro-inflammatory macrophages. However, when lactic acid and PA are used in combination, we observe a further increase in the percentage of CD86+ pro-inflammatory macrophages (Fig. 4G).

It is well-established that arginase-1 (Arg-1) and inducible nitric oxide synthase (iNOS) are hallmark molecules of anti-inflammatory and pro-inflammatory macrophages, respectively. Therefore, monitoring the expression of these enzymes in macrophages is crucial for a deeper understanding of their polarization states and functions. Our immunofluorescence data align with the results obtained from flow cytometry, demonstrating that lactic acid alone upregulated the anti-inflammatory marker Arg-1 in macrophages, while PA enhanced the expression of the pro-inflammatory marker iNOS. Furthermore, the combined stimulation of lactic acid and PA led to further upregulation of iNOS expression (Additional file 1: Fig. S4D).

Simultaneously, we explored changes in lactic acid levels and histone lactylation in RAW264.7 macrophages under these stimulatory conditions. The results indicated that lactic acid alone significantly elevated intracellular lactic acid levels, accompanied by enhanced lactylation at the H3K18 and H4K12 sites. In contrast, while PA alone does not exert a notable influence on intracellular lactic acid levels, it specifically and substantially promotes lactylation at the H4K12 site. Notably, when lactic acid and PA are combined, not only does the intracellular lactic acid level increase, but the lactylation at the H4K12 site also exhibits a further enhancement trend (Fig. 4H, Additional file 1: Fig. S4E). These results suggest that exogenous lactic acid may enhance PA-induced pro-inflammatory polarization of macrophages by mediating H4K12 lactylation.

Further investigation revealed that PA could upregulate the transcription levels of various inflammatory factors in macrophages, such as Tnf-α, Il-1β, Il-6, and Il-18 (Fig. 4I, Additional file 1: Fig. S4F). When combined with lactic acid, the transcription levels of Tnf-α and Il-1β were further amplified (Fig. 4I), confirming the role of lactic acid in enhancing PA-induced inflammatory responses in macrophages. Additionally, we observed a significant role for the hypoxia-inducible factor HIF-1α in this process, which is a crucial regulator of anti-inflammatory/pro-inflammatory polarization in macrophages. Our study found that under PA stimulation, the transcription level of Hif-1α increased, and when combined with lactic acid, the transcription level of Hif-1α was further elevated (Fig. 4J). By utilizing H4K12La antibodies to pull down protein-DNA complexes and performing de-crosslinking experiments, we discovered that the Hif-1α DNA content in the PA-treated group was significantly higher than that in the control group. The lactic acid and PA combination group had even higher Hif-1α DNA content. Notably, there was no significant difference in iNOS DNA content between the three groups (Fig. 4K). These findings suggest that lactic acid facilitates the transcription of Hif-1α by augmenting H4K12La levels, ultimately potentiating the inflammatory response induced by PA in macrophages.

VB124 improves cardiac injury and reduces inflammatory macrophage infiltration in type 2 diabetic mice

After a 4-week treatment with VB124, administered intraperitoneally at a dosage of 10 mg/kg/daily, a notable reduction in the left ventricular mass (LV mass) was observed in Leprdb mice, accompanied by an increase in both the E/A ratio and left ventricular ejection fraction (LVEF) as illustrated in Fig. 5A. These data unequivocally establish that VB124 can significantly benefit cardiac hypertrophy and diastolic and systolic functions in type 2 diabetic mice. Pathological staining further revealed that VB124 treatment substantially reduced myocardial hypertrophy, interstitial fibrosis, and lipid droplet deposition in the Leprdb mice (Fig. 5B, Additional file 1: Fig. S5A). This observation was further corroborated by RT-qPCR analysis, which demonstrated a significant downregulation of the transcription levels of Anp and Bnp, markers of myocardial injury in Leprdb mice, following VB124 treatment (Fig. 5C). Moreover, VB124 was also found to significantly decrease the transcription levels of Cd36 and fatty acid-binding protein 3 (Fabp3), key players in fatty acid transport (Fig. 5D), further highlighting the critical role of VB124 in ameliorating cardiac lipid droplet deposition.

Fig. 5

*Leprdb compared to m Leprdb, #Leprdb + VB124 compared to Leprdb. A Pulsed‐wave Doppler showing: LV mass, E/A ratio and LVEF%. B Representative diagram and quantification of cardiac hypertrophy, fibrosis and oil-drop deposition in mice. C mRNA expression of cardiac Anp and Bnp in mice. D mRNA expression of cardiac Cd36 and Fabp3 in mice. E mRNA expression of cardiac Il-1β and Ccl2 in mice. F Representative diagram and quantification of heart H4K12La positive macrophage (%) in mice. G Representative diagram and quantification of heart HIF-1α positive macrophage (%) in mice. H Representative diagram and quantification of heart iNOS positive macrophage (%) in mice

Importantly, our study also revealed that VB124 treatment led to a marked reduction in ROS levels in the hearts of Leprdb mice, concomitant with an increase in double-stranded DNA (dsDNA) content within mitochondrial compartments (Additional file 1: Fig. S5B). This provides direct evidence of improved mitochondrial function in the heart. Additionally, cytokine profiling revealed a significant downregulation of inflammatory mediators Il-1β and Ccl2 transcription levels in Leprdb mice receiving VB124 treatment (Fig. 5E), further corroborating the anti-inflammatory properties of VB124. Finally, through macrophage analysis, we observed a notable decrease in the infiltration of H4K12La, HIF-1α, iNOS, and IL-1β-positive macrophages in the hearts of Leprdb mice treated with VB124 (Fig. 5F–H, Additional file 1: Fig. S5D). This finding not only reinforces the anti-inflammatory effects of VB124 but also sheds light on a potential mechanism underlying its cardioprotective effects through modulation of macrophage activity.

Correlation between blood lactate concentration and cardiac injury in T2DM and its application in a predictive model for diastolic dysfunction

We gathered comprehensive information, medical histories, and medication usage from patients with clearly diagnosed T2DM. These patients were then divided into two groups based on their peripheral blood lactate concentrations: the normal lactate group (Lac < 2.2 mmol/L) and the elevated lactate group (Lac ≥ 2.2 mmol/L). A comparative analysis of the baseline data between these two groups and logistic regression analysis revealed gender to be a significant independent factor influencing hyperlactatemia in T2DM patients. Specifically, women were found to have a significantly lower risk of developing hyperlactatemia compared to men (Additional file 2: Tables S1, S2). Furthermore, an examination of laboratory test indicators and linear regression analysis identified blood glucose and triglycerides as independent factors influencing peripheral blood lactate concentrations, with a notably positive correlation observed (Additional file 2: Tables S3, S4). A more detailed comparison of cardiac observation indicators between the two groups, along with correlation analysis, highlighted a significant negative correlation between CK-MB, hs-TNT, NT-proBNP, MYO, LVDd, and the E/A ratio with peripheral blood lactate concentrations (Additional file 2: Table S6). These negative correlations remained consistent even after adjusting for confounding variables such as gender and diabetes duration (Additional file 2: Table S7).

Utilizing the aforementioned clinical dataset, we developed a clinical prediction nomogram specifically tailored for T2DM patients with cardiac diastolic dysfunction. This nomogram was constructed based on a training dataset comprising 599 T2DM patients and was externally validated using an independent dataset of 299 patients. A statistical analysis of the two datasets revealed no significant differences in most variable factors, except the proportion of sulfonylurea and insulin usage (Additional file 2: Table S8). A comprehensive statistical and logistic regression analysis of the general characteristic variables within the training dataset identified gender, age, alcohol consumption, systolic blood pressure, diastolic blood pressure, and lactate as independent predictors of T2DM with cardiac diastolic dysfunction (Additional file 2: Table S9). Leveraging these predictive indicators, we constructed a predictive nomogram for T2DM with cardiac diastolic dysfunction (Fig. 6A). This model demonstrated impressive discriminatory power, with an AUC (area under the ROC curve) of 0.7795 (95% CI 0.747–0.843), a sensitivity of 0.758, and a specificity of 0.748 (Fig. 6B). Additionally, our model exhibited strong goodness-of-fit, as evidenced by the close alignment of the nomogram’s calibration curve with the ideal value (Fig. 6C), and the decision curve analysis further highlighted its significant net benefit (Fig. 6D). Finally, we substantiated the reliability of our predictive model through rigorous external validation (Fig. 6E–G). These findings underscore the profound relationship between lactate and cardiac injury in T2DM and establish lactate as a robust independent predictor for assessing cardiac diastolic dysfunction. From a different perspective, this result reinforces the crucial role of regulating lactate transport in safeguarding the hearts of patients with T2DM from injury.

Fig. 6

A Nomogram for the prediction of diastolic function in patients with T2DM. B ROC curve of training set, ROC receiver operating characteristic, AUC area under the ROC curve. C Calibration curve for predicting probability of diastolic function in patients with T2DM in training set. D Decision curve analysis in prediction of diastolic function in patients with T2DM in training set. E ROC curve of validation set. F Calibration curve for predicting probability of diastolic function in patients with T2DM in validation set. G Decision curve analysis in prediction of diastolic function in patients with T2DM in the validation set