Chronic Hyperglycaemia Inhibits Tricarboxylic Acid Cycle in Rat Cardiomyoblasts Overexpressing Glucose Transporter Type 4

An oversupply of nutrients with a loss of metabolic flexibility and subsequent cardiac dysfunction are hallmarks of diabetic cardiomyopathy. Even if excess substrate is offered, the heart suffers energy depletion as metabolic fluxes are diminished. To study the effects of a high glucose supply, a stably glucose transporter type 4 (GLUT4)-overexpressing cell line presenting an onset of diabetic cardiomyopathy-like phenotype was established. Long-term hyperglycaemia effects were analysed. Rat cardiomyoblasts overexpressing GLUT4 (H9C2KE2) were cultured under normo- and hyperglycaemic conditions for long-term. Expression profiles of several proteins were compared to non-transfected H9C2 cells (H9C2) using RT-qPCR, proteomics-based analysis, or Western blotting. GLUT4 surface analysis, glucose uptake, and cell morphology changes as well as apoptosis/necrosis measurements were performed using flow cytometry. Additionally, brain natriuretic peptide (BNP) levels, reactive oxygen species (ROS) formation, glucose consumption, and lactate production were quantified. Long-term hyperglycaemia in H9C2KE2 cells induced increased GLUT4 presence on the cell surface and was associated with exaggerated glucose influx and lactate production. On the metabolic level, hyperglycaemia affected the tricarboxylic acid (TCA) cycle with accumulation of fumarate. This was associated with increased BNP-levels, oxidative stress, and lower antioxidant response, resulting in pronounced apoptosis and necrosis. Chronic glucose overload in cardiomyoblasts induced by GLUT4 overexpression and hyperglycaemia resulted in metabolically stimulated proteome profile changes and metabolic alterations on the TCA level.


Introduction
Diabetic cardiomyopathy (DC) is characterised by marked structural and morphological changes at the myocardial level, resulting in left ventricular dysfunction [1]. It develops independently from further cardiac risk factors such as hypertension or coronary artery disease that progresses to heart failure (HF) as a secondary manifestation of diabetes [1]. In general, this patient group has a high mortality, corresponding to 1% of the cases among all patients diagnosed with diabetes [2,3]. In this context, diabetes mellitus per se is a cardiovascular risk factor and multiplies the risk for heart failure [3]. Although the mechanisms leading to DC are not fully understood, it is well-accepted that the combination of hyperglycaemia, reactive glucose metabolites such as methylglyoxal, and increased formation of advanced glycation end products (AGEs) in cardiomyocytes and endothelial cells contribute to the development of the disease [2,4]. AGEs stimulate structural, functional, and transcriptional changes. As a consequence of glucose and glycation overflow, the detoxifying enzymes progressively fail, limiting the glyoxalase function in the methylglyoxal elimination process. Insufficient activation of endogenous repair mechanisms results in mitochondrial dysfunction, increased glycation, and oxidative stress [5]. The combination of hyperglycaemic state and increased methylglyoxal concentration prolongs the glucose transporter presence at the cell surface and increases glucose influx, which is insulin-independent [6,7]. This state triggers a vicious cycle of glucose intoxication with subsequent glycation [6,7], resulting in oxidative and mitochondrial stress [8]. Glucose overflow induces impaired transcription of metabolic enzymes and structural molecules [6,7] and alters the mitochondrial function by post-translational modification [8]. These findings suggest an important role of glucose overflow in metabolic-induced organ damage.
In addition, and predominantly in later stages of metabolic impairment, insulin resistance supports a metabolic shift increasing the β-oxidation, intracellular fatty acid concentration, and reactive oxygen species production [2]. These events contribute to cardiomyocyte malfunction and death, cardiac hypertrophy, and fibrosis [2]. At this level, further alteration at Ca 2+ signalling and titin phosphorylation are also observed, providing a beneficial scenario for the development of HF [2].
Another important metabolic pathway possibly associated with DC development is the tricarboxylic acid (TCA) cycle [9]. This cycle is responsible for producing reduced nicotinamide adenine dinucleotide (NADH) through the oxidation of acetyl-coenzyme A (acetyl-CoA) derived from carbohydrates, fatty acids, and proteins [10]. In type 1 diabetes mellitus, this cycle is highly altered, contributing to the development of cardiovascular autonomic neuropathy [11]. A similar effect was observed in type 2 diabetes mellitus, resulting in an inefficient mitochondrial TCA cycle influx and reduced energy yield in skeletal muscle [12]. Following this dysfunction, high levels of insulin lead to lower expression of citrate synthase, one of the first enzymes responsible for initiating the TCA cycle activity [12]. This reduction in the diabetic mitochondria influx is based on profound posttranscriptional modifications of all TCA cycle enzymes [13].
Little is known about the TCA cycle response in DC. Particularly, full changes at the proteome levels or the gene and protein expression of these enzymes in diabetic myocytes have not been investigated so far. One recent paper proved that myocardial glucose utilisation is critical in diabetic cardiomyopathy. Restoring glucose uptake was associated with impairments of mitochondrial energy gain, exerting new effects of glucose in diabetic heart metabolism [8]. To further address this, the present study aimed to develop a suitable model for reproducing the onset of DC. A model that could mimic the hyperglycaemic state, the overexpression of glucose transporter type 4 (GLUT4), and, by glucose overload, the inactivation of the glyoxalase system in cardiomyocytes.
Therefore, we established a rat cardiomyoblast cell line which stably overexpressed GLUT4 (H9C2KE2). These cells were exposed to hyperglycaemia for nine months to develop a DC-onset-like phenotype. Additionally, metabolic alterations were monitored concerning the TCA cycle under increased glucose influx in cardiomyoblasts.

Overexpression of GLUT4 in Hyperglycaemic H9C2 as a Possible Model for Mimicking the Onset of DC
To measure whether the overexpression of GLUT4 in H9C2 can be a suitable model for mimicking the onset of DC, cells were chronically exposed to 20 mM or 30 mM glucose for nine months (long-term exposition, L). These glucose concentrations resemble normoand hyperglycaemia in this cell type [14]. After this time, the GLUT4 presence on the cell surface significantly increased in normo-and hyperglycaemic H9C2KE2 cells if compared to wild type cells (2-fold more, WT30L vs. KE230L, p < 0.0001/1.4-fold more, WT20L vs. KE220L, p < 0.0001, t-test) ( Figure 1A). However, in terms of gene expression and protein . Representative Western blots are also shown. Data are shown as mean ± SD values described as WT20L vs. KE220L and WT30L vs. KE230L with * p < 0.05, ** p < 0.01, *** p < 0.001, or **** p < 0.0001 Representative Western blots are also shown. Data are shown as mean ± SD values described as WT20L vs. KE220L and WT30L vs. KE230L with * p < 0.05, ** p < 0.01, *** p < 0.001, or **** p < 0.0001 for showing the significance. Shapiro-Wilk test was applied for testing the data normality distribution. For t-test analysis, an unpaired test for all normally distributed data and Mann-Whitney for all abnormally distributed data was used. (WTL-Wild-type cells exposed for long-term/KE2L-GLUT4 overexpressing cells exposed for long-term).
The overexpression of GLUT4 markedly influenced the enzymatic activity of glyoxalase 1 (GLO1) after chronic exposure to glucose (regardless of the concentration) ( Figure 1F) (1.6-fold more, WT20L vs. KE220L, p < 0.0001/2.3-fold more, WT30L vs. KE230L, p < 0.0001, t-test). However, a similar increment was not detected in the protein expression and gene expression. The GLO1 gene expression showed a gradual reduction when compared to the WT cells (1.5-fold less, WT20L vs. KE220L, p < 0.0001/1.3-fold less, WT30L vs. KE230L, p < 0.05, one-way ANOVA), followed by no alterations at the protein expression levels ( Figure S1B,C, respectively). Moderate changes were found in terms of D-lactate production ( Figure S1D). However, this alteration was just effective when normoglycaemic H9C2KE2 cells were compared with wild-type cells (3-fold, p < 0.001, one-way ANOVA).

Glucose Overflow Might Inhibit the Last Steps of the TCA Cycle
A proteomic-based analysis was performed, quantifying a total of 2654 proteins. Among these quantified proteins, 21 (from 22) TCA cycle annotated proteins could be detected. To estimate their absolute abundance in the different sample groups, we used iBAQ, the intensity-based absolute quantification option provided in MaxQuant (Table S2).
Key regulatory enzymes were analysed further ( Figure 2). The abundance of pyruvate carboxylase was slightly higher in hyperglycaemic H9C2KE2 cells when compared with hyperglycaemic WT cells. However, in all H9C2KE2 cells, a high abundance of pyruvate carboxylase could be observed regardless of the level of glucose, compared to WT cells. The abundance of fumarase was consistently lower in H9C2KE2 cells, regardless of the level of glucose, when compared with WT cells. However, succinate dehydrogenase presented with lower abundance in hyperglycaemic H9C2KE2 cells when compared with hyperglycaemic WT cells.
In terms of enzymatic activity, we found that fumarate concentrations were significantly increased, particularly after chronical hyperglycaemic in H9C2KE2 cells, when compared with WT cells (1.22-fold less, WT20L vs. KE220L, p < 0.01/1.9-fold more, WT30L vs. KE230L, p < 0.001, t-test) ( Figure 3D). with hyperglycaemic WT cells. However, in all H9C2KE2 cells, a high abundance of pyruvate carboxylase could be observed regardless of the level of glucose, compared to WT cells. The abundance of fumarase was consistently lower in H9C2KE2 cells, regardless of the level of glucose, when compared with WT cells. However, succinate dehydrogenase presented with lower abundance in hyperglycaemic H9C2KE2 cells when compared with hyperglycaemic WT cells. shown as mean ± SD values (WTL-Wild-type cells exposed for long-term/KE2L-GLUT4 overexpressing cells exposed for long-term).
The protein expression of pyruvate carboxylase followed the same profile as observed in fumarase, when compared with WT cells (1.27-fold less, WT20L vs. KE220L, p < 0.05/1.39-fold less, WT30L vs. KE230L, p < 0.05, t-test) ( Figure 3B). Succinate dehydrogenase B was unaffected in chronically hyperglycaemic H9C2KE2 cells when compared with Figure 2. Proteomic changes in H9C2KE2 cells at the TCA cycle level including PDH-reaction. IBAQ values represent the relative abundance of detected proteins within the respective analysed set. Pdha1 = pyruvate dehydrogenase complex A, Pdhb = Pyruvate dehydrogenase complex B, Sdhb = succinate dehydrogenase complex B, Fh = fumarase, Pc = pyruvate carboxylase, N = 3, data are shown as mean ± SD values (WTL-Wild-type cells exposed for long-term/KE2L-GLUT4 overexpressing cells exposed for long-term).

GLUT4 Overexpression Followed by Hyperglycaemia-Effects on Glycolysis
Following the enzymatic evaluation, the next step was to closely control proteins associated with these enzymes, particularly proteins related to glycolysis. We found that GAPDH was upregulated at the gene expression level in hyperglycaemic H9C2KE2 cells, particularly when compared among the glucose concentrations (1.4-fold more, WT30L vs. KE230L, p < 0.05/1.4-fold more, KE220L vs. KE230L, p < 0.05, one-way ANOVA) ( Figure S2A). However, this regulation did not extrapolate into protein expression level ( Figure 4A).
In terms of enzymatic activity, we found that fumarate concentrations were signifi-cantly increased, particularly after chronical hyperglycaemic in H9C2KE2 cells, when compared with WT cells (1.22-fold less, WT20L vs. KE220L, p < 0.01/1.9-fold more, WT30L vs. KE230L, p < 0.001, t-test) ( Figure 3D). . Representative Western blots are also shown. Data are shown as mean ± SD values described as KE220L vs. KE230L and WT30L vs. KE230L with * p < 0.05, **** p < 0.0001 for showing the significance. A Shapiro-Wilk test was applied for testing the data normality distribution. For t-test analysis, an unpaired test for all normally distributed data and a Mann-Whitney for all abnormally distributed data was used. (WTL-Wild-type cells exposed for long-term/KE2L-GLUT4 overexpressing cells exposed for long-term). . Representative Western blots are also shown. Data are shown as mean ± SD values described as KE220L vs. KE230L and WT30L vs. KE230L with * p < 0.05, **** p < 0.0001 for showing the significance. A Shapiro-Wilk test was applied for testing the data normality distribution. For t-test analysis, an unpaired test for all normally distributed data and a Mann-Whitney for all abnormally distributed data was used. (WTL-Wild-type cells exposed for long-term/KE2L-GLUT4 overexpressing cells exposed for long-term). phorylation was significantly different between culture conditions. In hyperglycaemia, increased phosphorylation occurred: (WT20L vs. WT30L, p < 0.05; KE220L vs. KE230L, p < 0.01, t-test) but phosphorylation grade was independent of cell type ( Figure 4C,D).
In terms of pyruvate, this increased production persisted only when comparing hyperglycaemic H9C2KE2 cells with hyperglycaemic WT cells (1.6-fold more, WT30L vs. KE230L, p < 0.01, t-test) after the long-term exposition ( Figure 4E). . Representative Western blots are also shown. Data are shown as mean ± SD values described as KE220L vs. KE230L, WT20L vs. KE220L, and WT30L vs. KE230L with * p < 0.05, ** p < 0.01, or *** p < 0.001 for showing the significance. Shapiro-Wilk test was applied for testing the data normality distribution. For t-test analysis, an unpaired test for all normally distributed data and a Mann-Whitney for all abnormally distributed data was used. (WTL-Wild-type cells exposed for long-term/KE2L-GLUT4 overexpressing cells exposed for long-term).

The Accumulation of Fumarate in H9C2KE2 Cells Induces Oxidative Stress
Fumarate accumulation in H9C2KE2 cells resulted in an upregulation of NRF2 gene expression when compared with WT cells (1.3-fold more, WT20 vs. KE220, p < 0.01/1.5-fold more, WT30L vs. KE230L, p < 0.0001, one-way ANOVA), and was even more pronounced when compared between the hyper-and normoglycaemic H9C2KE2 cells (2.15-fold more, p < 0.0001, one-way ANOVA) ( Figure S3A). However, only the phosphorylation of Nrf2 was upregulated by long-term hyperglycaemia when compared with WT cells (2-fold more, WT30L vs. KE230L, p < 0.001, t-test) or among to the H9C2KE2 cells (p < 0.05) ( Figure 5A,B). These changes were not extended to KEAP1 (WT20L vs. KE220L, p > 0.05/1.4-fold more, WT30L vs. KE230L, p < 0.05, one-way ANOVA), showing that the gene expression of this protein was not able to completely inhibit the Nrf2 activity ( Figure S3B). shown. Data are shown as mean ± SD values de-scribed as KE220L vs. KE230L, WT20L vs. KE202L, and WT30L vs. KE230L with * p < 0.05, ** p < 0.01, *** p < 0.001, or **** p < 0.0001 for showing the significance. Shapiro-Wilk test was applied for testing the data normality distribution. For t-test analysis, an unpaired test for all normally distributed data and a Mann-Whitney for all abnormally distributed data was used. (WTL-Wild-type cells exposed for long-term/KE2L-GLUT4 overexpressing cells exposed for long-term).
The profound changes induced by overexpression of GLUT4 associated with hyperglycaemia could no longer avoid the activation of cell death mechanisms, proved by increased levels of apoptosis (1.7-fold more, WT30L vs. KE230L, p < 0.0001/1.6-fold more, KE220L vs. KE230L, p < 0.0001, one-way ANOVA) ( Figure 6B). However, in the case of necrosis, this alteration is more likely related to hyperglycaemia than to GLUT4 overexpression itself, as hyperglycaemic H9C2KE2 and WT cells showed a significant increase of necrosis events when compared with the same cells under normoglycaemic levels (2fold more, WT20L vs. WT30L, p < 0.0001; 2.25-fold more, KE220L vs. KE230L, p < 0.0001, . Representative Western blots are also shown. Data are shown as mean ± SD values de-scribed as KE220L vs. KE230L, WT20L vs. KE202L, and WT30L vs. KE230L with * p < 0.05, ** p < 0.01, *** p < 0.001, or **** p < 0.0001 for showing the significance. Shapiro-Wilk test was applied for testing the data normality distribution. For t-test analysis, an unpaired test for all normally distributed data and a Mann-Whitney for all abnormally distributed data was used. (WTL-Wild-type cells exposed for long-term/KE2L-GLUT4 overexpressing cells exposed for long-term).
The protein expression of PGC1α was lower in hyperglycaemic cells when compared with normoglycaemic H9C2KE2 cells (1.25-fold less, KE220L vs. KE230L, p < 0.01, oneway ANOVA) but was higher when compared with WT cells. This might occur because normoglycaemic H9C2KE2 cells induce a significant overexpression of this protein (p < 0.01, when compared with normoglycaemic WT cells). It is expected that H9C2KE2 cells lose antioxidant protection upon stimulation with high levels of glucose ( Figure 5C).

The Accumulation of Fumarate in H9C2KE2 Cells Activates Cell Death
Hyperglycaemic H9C2KE2 cells showed significant overexpression of BNP when compared with hyperglycaemic WT cells (8.3-fold more, WT30L vs. KE230L, p < 0.0001) or among the H9C2KE2 cells (2.5-fold more, p < 0.0001, one-way ANOVA) ( Figure 6A). one-way ANOVA) ( Figure 6C). However, only normoglycaemic KE2 cells had higher events when compared with normoglycaemic WT cells (1.2-fold more, WT20L vs. KE220L, p < 0.01; t-test). Shapiro-Wilk test was applied for testing the data normality distribution. For t-test analysis, an unpaired test for all normally distributed data and a Mann-Whitney for all abnormally distributed data was used. (WTL-Wild-type cells exposed for long-term/KE2L-GLUT4 overexpressing cells exposed for long-term).
Therefore, considering all the structural, metabolic, and molecular changes induced by overexpression of GLUT4 in H9C2KE2 cells exposed to glucose, we have a basis for indicating that H9C2KE2 can be a suitable model for mimicking the metabolically induced cardiomyopathy. Data are shown as mean ± SD values described as KE220L vs. KE230L, WT20L vs. KE220L, and WT30L vs. KE230L with * p < 0.05, ** p < 0.01, or **** p < 0.0001 for showing the significance (N = 3). Shapiro-Wilk test was applied for testing the data normality distribution. For t-test analysis, an unpaired test for all normally distributed data and a Mann-Whitney for all abnormally distributed data was used. (WTL-Wild-type cells exposed for long-term/KE2L-GLUT4 overexpressing cells exposed for long-term).
The profound changes induced by overexpression of GLUT4 associated with hyperglycaemia could no longer avoid the activation of cell death mechanisms, proved by increased levels of apoptosis (1.7-fold more, WT30L vs. KE230L, p < 0.0001/1.6-fold more, KE220L vs. KE230L, p < 0.0001, one-way ANOVA) ( Figure 6B). However, in the case of necrosis, this alteration is more likely related to hyperglycaemia than to GLUT4 overexpression itself, as hyperglycaemic H9C2KE2 and WT cells showed a significant increase of necrosis events when compared with the same cells under normoglycaemic levels (2-fold more, WT20L vs. WT30L, p < 0.0001; 2.25-fold more, KE220L vs. KE230L, p < 0.0001, one-way ANOVA) ( Figure 6C). However, only normoglycaemic KE2 cells had higher events when compared with normoglycaemic WT cells (1.2-fold more, WT20L vs. KE220L, p < 0.01; t-test).
Therefore, considering all the structural, metabolic, and molecular changes induced by overexpression of GLUT4 in H9C2KE2 cells exposed to glucose, we have a basis for indicating that H9C2KE2 can be a suitable model for mimicking the metabolically induced cardiomyopathy.

Discussion
In H9C2 cardiomyoblasts, chronic hyperglycaemia leads to increased glucose influx. In our model of GLUT4 overexpressing H9C2 cells, this is even more pronounced, although GLUT4 expression per se is downregulated on mRNA as well as on the protein level. Prolonged presentation of GLUT4 on the cell surface by reduced internalisation was detected and might explain increased glucose influx and consumption as shown in our model. Reactive glucose metabolites originating from glycolysis overflow induce increased detoxifying mechanisms, as shown by increased glyoxalase activity. These findings suggest a direct impact of glucose influx on cardiomyocyte fate, as it is associated with generation of oxidative stress and subsequent cell death. This effect was insulin-independent, suggesting deleterious cardiac effects beyond the classical model of insulin resistance as being an exclusive driver of DC. Therefore, we assume that our model may mimic the early phase in which surplus of nutrients and subsequent glycation boost the onset of the disease.
Previously we have demonstrated a deleterious effect of glucose overflow in L6 myoblasts and identified the reactive dicarbonyl methylglyoxal as being involved in this pathology. Methylglyoxal stress prolonged GLUT4 presence at the cell surface of myoblasts [6]. This causes exaggerated glucose influx and high systemic glucose availability, resulting in activation of cell death mechanisms [6]. To analyse whether this myoblastderived hypothesis is valuable in cardiomyoblasts, the well-established H9C2 cell culture model was adapted for this purpose [14]. GLUT4-overexpressing H9C2 cells (H9C2KE2) were developed to study the increased glucose uptake. By inducing this high-glucose influx, we assume increased intracellular methylglyoxal (MG) production from overflow in glycolysis, as shown before in L6 myoblasts [6]. H9C2KE2 cells react upon methylglyoxal stress via direct incubation with methylglyoxal with increased glucose uptake in a similar amount as if treated with insulin. This previous observation adds a new viewpoint to early pathologic mechanisms resulting in diabetes-induced cardiomyocyte damage and the role of reactive glucose metabolites in this scenario [15]. Increased glucose uptake induced by hyperglycaemia leads to impairments in glycolysis, generating reactive glucose metabolites such as methylglyoxal which then prolong GLUT4 presence on the cell surface, allowing for even more glucose to enter the cell. This is a self-accelerating vicious cycle leading to glucotoxicity influencing in a second instance metabolic pathways for energy generation, as proven by our data. Therefore, the classical view of cardiomyopathy as being induced by insulin resistance and low uptake of metabolites may be questioned in hyperglycaemia; besides the amount of GLUT4, its functionality and surface presentation should be recognised as being involved in the pathogenesis of a metabolically induced cardiomyocyte dysfunction, at least in early stages.
The morphological changes observed in H9C2KE2 cells also indicated that this model could mimic the structural changes associated with the onset of DC. These cells showed increased size and granularity as a result of enhanced protein synthesis and turnover.
Long-term effects of high glucose and fatty acids on cardiomyocytes result in hypertrophy, particularly via upregulation of atrial natriuretic peptide (ANP), connective tissue growth factor, and α-skeletal actin gene expression [16]. BNP levels are a primary marker of cardiac dysfunction [17][18][19], especially because BNP activates lipid peroxidation and apoptosis, which are important inducers of heart failure at the level of the myocardium. Interestingly, the increase in the levels of BNP in DC is not fully understood, considering that most of these patients develop heart failure from the diabetic condition instead of having previous cardiac dysfunction [20]. Hence, we were the first to study an increased level of BNP under such conditions, proving that hyperglycaemic H9C2KE2 cells have a disturbed myocardial performance.
Glucose overload by GLUT4 overexpression can worsen the mitochondrial respiratory chain capacity of myocardial cells, yielding less efficient energy production. This effect was mainly driven by glycosylation of the transcription factor specificity protein 1 (SP1), a is master regulator of mitochondrial protein expression, identifying mitochondria as a major target of glucotoxicity [8]. To answer the question of whether increased glucose influx affects key metabolic pathways, we designed our GLUT4-overexpressing H9C2 cells based on previous experiences with GLUT4-overexpressing L6 myoblasts by using a protocol adapted from the working group of Amira Klip [21]. The negative effect on GLUT4 cycling that was observed in L6-cells was similarly reproducible in H9C2 cells, suggesting a generalisable pathology associated with GLUT4 traffic that might be constrained by reactive glucose metabolites. Based on this, H9C2KE2 is a useful model for studying the effects of chronic glucose overflow and glucotoxicity on the myocardium.
It is known that extracellular stimulation of methylglyoxal or knockdown levels of glyoxalase 1 raises the methylglyoxal concentrations, inducing the translocation of GLUT4 [6,7]. Similarly, in our model, reduced transcription of glyoxalase 1 was found under high GLUT4 presence on the surface, although higher GLO1 enzyme activity, as well as increased Nrf2 expression and phosphorylation, were detected. This "transient positive" response might be explained by the relevant strength of the detox system in reaching exhaustion, the strong supportive mechanism between the glyoxalase and the antioxidant enzymes, and the limited power of our model to stimulate the glycation process. We assume that these beneficial effects tend to disappear after chronic glucose influx, driving the glyoxalase and antioxidant systems to a breakdown, as seen in long-standing diabetes patients and at the myocardium level under huge pressure overload [22]. This conclusion is also supported by the fact that increased levels of apoptosis and necrosis, particularly in hyperglycaemic H9C2KE2 cells, were already observed in our model.
Following the metabolic changes, the TCA cycle was affected on the level of fumarate metabolism in our model. Consequently, the total protein responses were evaluated in order to understand on a more complex level how TCA cycles enzymes demand for keeping its function. However, an opposite regulation for some proteins between proteomics and Western blotting was found.
In the hypoxia model, the accumulation of several TCA compounds was detected by metabolomic analysis [26,27]. Whereas, particularly in the ischaemia model, the accumulation of succinate was described as a disease-specific signature, being responsible for mitochondrial ROS production during reperfusion [28]. This latter study also proved that succinate accumulation was influenced by fumarate overflow, which was probably in charge of reversing the succinate dehydrogenase reaction [28]. This fumarate overflow was suspected to arise from purine nucleotide metabolism and from the partial reversal of the malate/aspartate shuttle in this scenario. Therefore, during reperfusion, re-oxidation of succinate generates ROS in the mitochondrial complex I [28]. Together, these findings support the dual role of fumarate in inducing ROS and stimulating antioxidative pathways via Nrf2-stimulation.
Similarly, we observed increased ROS formation and ROS-associated proteins as a possible result of an inefficient TCA cycle activity. Our possible explanation is that the accumulation of fumarate concomitantly suppresses the activity of succinate dehydrogenase, generating even more ROS as a vicious cycle. As a consequence, pyruvate carboxylase (PC) fails to recompense the oxaloacetate loss in the TCA cycle, due to the increased glucose influx (in H9C2KE2 cells). Then, PC is downregulated to nearly 50% when compared with H9C2KE2 normoglycaemic cells. Besides this, pyruvate dehydrogenase complex was upregulated and, via increased phosphorylation, blocked in its activity, preventing the processing of pyruvate to acetyl-CoA. This results in pyruvate accumulation, as shown by our measurements. However, in a normal condition, WT cells can restore the pyruvate levels and refill the TCA cycle, except for H9C2KE cells, which, at this level, already lost the restoring capacity. Thus, the TCA cycle of H9C2KE cells runs out of compounds. It was not our intention to analyse the effect of TCA agonists/antagonists within this experimental setting; however, one might assume that, besides tight glucose control, TCA activation might reduce fumarate levels and thus reduce symptoms of oxidative stress.
Interestingly, this is the first study that deeply investigated levels of proteins from the TCA cycle in a cell culture model of DC onset. Therefore, our findings indicate the need for improving glycaemic control as a basis for the prevention of the disease. Particularly in the myocardium, impaired GLUT4 traffic induced by reactive glucose metabolites and glucoseinduced loss of function of the TCA cycle can be observed as an essential mechanism. Before this present study, a clinical study was able to detect a high level of TCA metabolites in the plasma of patients with a high risk of heart failure; however, diabetes was not the only risk factor of these patients [29]. In addition, after knowing how the TCA cycle responds during long-term hyperglycaemia, future studies can better address the therapy based on intermediate metabolites from the TCA cycle for treating DC.

Conclusion
In the classical view of DC, the myocardium suffers from a lack of energy although and due to the surplus of both glucose and fatty acids. Both compounds hamper each other's metabolism, and thus the heart runs out of fuel [30]. In summary, our findings demask how the metabolic stress induced by hyperglycaemia can affect the TCA cycle at the fumarase level, resulting in a probable loss of NADH. Hereby, at the level of heart cells, the TCA cycle may run out of components that are necessary for functioning, resulting in inefficient energy production and a lack of fuel (Figure 7). Our model might resemble the beginning of the process of DC, induced by glucose overflow as happens in diabetes mellitus at early stages. As the observed effects are insulin-independent, our model might apply for both types of diabetes, type 1 and type 2, providing a more prospective opportunity for the development of new therapies. cells, the TCA cycle may run out of components that are necessary for functioning, resulting in inefficient energy production and a lack of fuel (Figure 7). Our model might resemble the beginning of the process of DC, induced by glucose overflow as happens in diabetes mellitus at early stages. As the observed effects are insulin-independent, our model might apply for both types of diabetes, type 1 and type 2, providing a more prospective opportunity for the development of new therapies. Figure 7. Overview on metabolic stress in H9C2KE2 cells induced by glucose overflow; delivered products in green, demanded products in red. Within the TCA cycle, reduced amounts of fumarase induce fumarate overflow and lack of malate and following compounds of the cycle such as oxaloacetate. Oxaloacetate might be refilled by the pyruvate carboxylase at the expense of ATP. With the TCA cycle running out of substrates, the respiratory chain is lacking protons used for ATP production. Accumulation of fumarate is inducing oxidative stress via blocking the succinate Figure 7. Overview on metabolic stress in H9C2KE2 cells induced by glucose overflow; delivered products in green, demanded products in red. Within the TCA cycle, reduced amounts of fumarase induce fumarate overflow and lack of malate and following compounds of the cycle such as oxaloacetate. Oxaloacetate might be refilled by the pyruvate carboxylase at the expense of ATP. With the TCA cycle running out of substrates, the respiratory chain is lacking protons used for ATP production. Accumulation of fumarate is inducing oxidative stress via blocking the succinate dehydrogenase. Anaerobic glycolysis is pronounced as detected by increased L-lactate production. As a consequence of energy deprivation from the TCA cycle, increased BNP levels are released on the heart, provoking diastolic dysfunction. Observed findings are depicted with red arrows or lines, estimated consequences are shown in pink. The figure was drawn using pictures from "Servier Medical Art" (http://smart.servier.com/; accessed on 15 April 2022).

Hyperglycaemic Stage Measurements
Rat cardiomyocytes stably overexpressing GLUT4 (H9C2KE2) were exposed to normoglycaemic (20 mM glucose) and to hyperglycaemic (30 mM glucose) culture conditions for the long-term (9 months). After this time, the glucose uptake was determined by measuring the uptake of 2-(N- In parallel, glucose consumption and lactate production of each group of cells were also measured in cell culture supernatant via blood analyser (ABL800 FLEX radiometer, Krefeld, Germany).
During this time, the morphology of the cells was determined using an inverted microscopy (Eclipse TE2000-U, Nikon, Amsterdam, The Netherlands). The granularity and size were measured by flow cytometry (FC500, Beckman Coulter, Krefeld, Germany) and the total protein quantification was measured by bicinchoninic acid assay (cat. No. #B9643, Sigma-Aldrich, St. Louis, MI, USA).

Glyoxalase-1 Activity Assay
Glo1 activity was measured in H9C2KE2 cells according to Arai et al. [32]. Briefly, cells were washed with PBS and incubated for 3 min at 37 • C with trypsin-EDTA solution (cat.

Mass Spectrometry
Cell lysis was carried out as described using approx. 1-2 × 10 6 cells (>2 µg of protein) [33]. Cell pellets were resuspended in Urea buffer (30 mM TrisBase, 2 M Thiourea, 7 M Urea). For mechanical lysis, glass beads (0.25-0.5 mm and 1.25-1.65 mm) were added. Lysates were sonicated in a sonication block (Hielscher Ultrasonics GmbH, Teltow, Germany), with an amplitude of 90 and a cycle of 0.5 5× for 50 s, with rest on ice for 90 s in between cycles [34]. The resulting lysate was centrifuged (16000× g, 4 • C, 10 min) and the supernatant was transferred to another reaction tube. Protein concentration was determined via Bradford assay. Digestion of proteins into peptides was carried out. In brief, 10 µg of lysate were added to the digestion buffer (50 mM ammoniumbicarbonate). Samples were reduced with 15 mM dithiothreitol for 30 min at 56 • C and alkylated with 5 mM iodacetamide for 30 min in the dark at room temperature. Digestion was carried out over night at 37 • C, using trypsin as protease (ration 1:50). Digestion was stopped by acidification. Prior to mass spectrometry, the peptide concentration was determined by amino acid analysis and 200 ng of peptides were taken for the measurements [34]. Mass spectrometry was carried out as described by Plum et al. [35]. Briefly, nanoHPLC analysis was performed on an UltiMate 3000 RSLC nano LC system (ThermoFisher Scientific, Bremen, Germany). Peptides were loaded on a capillary pre-column (ThermoFisher Scientific, Bremen, Germany, 100 µm × 2 cm, particle size 5 µm, pore size 100 Å) and subsequently onto an analytical C18 column (ThermoFisher Scientific, Bremen, Germany, 75 µm × 50 cm, particle size 2 µm, pore size 100 Å). Peptide separation was achieved with a flow rate of 400 nL/min and a linear gradient up to 40% buffer B (84% acetonitrile, 0.1% formic acid). The HPLC system was online coupled to the nano ESI source of an Orbitrap Elite mass spectrometer (ThermoFisher Scientific, Bremen, Germany). The MS1 scan range was set from 300 to 2000 m/z and a resolution of 30,000. From each full scan, the Top 20 ions were selected for collision-induced dissociation (CID) fragmentation, with a normalised collision energy of 35%. The dynamic exclusion was set for 30 s. The subsequent data analysis was carried out using MaxQuant (v.1.6.10.43) [36]. The integrated Andromeda algorithm was taken to search spectra against the Uniprot rattus norvegicus reference proteome (11_2020) (UniProt, 2021) using trypsin as protease. The false discovery rate (FDR) was set to 1% for peptides (minimum length of 7 amino acids) and proteins and was determined by searching against a reverse decoy database. A maximum of two missed cleavages were allowed in the database search. Peptide identification was performed with an allowed initial precursor mass deviation up to 7 ppm and an allowed fragment mass deviation of 20 ppm. Carbamidomethylation of cysteines was set as fixed modification and oxidation of methionine as variable modification, due to sample pre-processing. Quantification was carried out using the MaxQuant Label Free Quantification (LFQ) algorithm including unique and razor peptides for quantification. For further quantification, the calculation of iBAQ values was enabled [37]. Resulting data were subsequently statistically analysed using Perseus (v. 1.6.14.0) [36]. iBAQ values were normalised by calculating the sum of all iBAQ values for each sample separately. To improve readability, iBAQ sums were divided by 1000. Single iBAQ values were subsequently divided by the respective sum (sample-wise). Resulting normalised iBAQ values were averaged for each sample group and were used for intensity-based absolute quantification of proteins. was calculated and normalised to protein concentration. The control group was assigned a value of 1, and the hyperglycaemic group was then calculated relative to the control group [40].

Brain Natriuretic Peptide (BNP) Measurement
Cells were centrifugated at 1500 rpm for 10 min at 4 • C and the supernatant was collected and quantified following the manufacturer's protocol from the BNP-32 rat ELISA kit (cat. No. #108815, Abcam, Cambridge, UK). A microplate reader (SunriseTM, Tecan, Maennedorf, Switzerland) was used for reading the samples with a wavelength of 450 nm. Correction of optical imperfections by subtracting readings at 570 nm was performed. After the measurements, the results were obtained based on the standard curve values and the multiplication of the dilution factor.

Statistical Analysis
Results of the experimental studies are reported as mean ± SD. Group comparisons among glucose concentration and cell type were analysed by one-way ANOVA followed by Tukey's multiple comparison post-test. A Shapiro-Wilk test was applied for checking the normality distribution. Direct comparisons between KE2 and WT cells in different conditions were analysed by unpaired t-test or Mann-Whitney test, respectively. Namely, for protein expression of GLUT4, glucose cell consumption, fumarase levels, protein expression of GAPDH, protein expression of PDH1α, protein expression of Nrf2, and protein expression of PGC1α, an unpaired t-test for normally distributed data was applied. For GLUT4 surface, NBDG levels, GLO1 activity, pyruvate carboxylase and pyruvate levels, SDHB, fumarate, pNrf2, ROS, apoptosis, and necrosis, a Mann-Whitney test for abnormally distributed data was applied. All of these analyses were performed using GraphPad Prism version 9.0.0 for Windows (GraphPad Software, San Diego, CA, USA). p-values < 0.05 were considered as statistically significant. N describes independent biological experiments.