Background
With the changing lifestyles, the incidence of diabetes mellitus (DM) shows a rapidly increasing trend. As estimated by the International Diabetes Federation, the number of DM patients will be increased to 0.5784 billion by 2030, resulting in a morbidity of up to 10.2% [
1]. DM increases the risk of developing heart failure (HF) by 2–4 times, as compared to healthy people [
2], and thus it tends to cause a highly poor prognosis. Diabetic cardiomyopathy (DCM) is one of the severe cardiovascular complications [
3] of DM first reported by S Rubler in 1972 [
4]. It is multi-factorial in pathophysiology and has not yet been fully explored.
Increasing studies have noted that mitochondrial events that lead to damage and dysfunction, including abnormal dynamics [
5], mitophagy [
6‐
8], calcium homeostasis imbalance [
9‐
11], disturbed energy metabolism and oxidative stress [
12,
13], play an essential role in DCM. In addition, excessive accumulation of lipid intermediary metabolites is considered as directly linked to the toxic injury and dysfunction of diabetic myocardium [
14]. It has been established that the immune infiltration and activation of inflammatory processes in myocardial tissues are also critical pathogeneses of DCM. For example, both type 1 and 2 DM models had increased myocardial infiltration of monocytes and macrophages [
15,
16]; chronic hyperglycemia induced elevation of Th1, Th2, and Th17 cytokines by activating T cells via the RAGE-dependent pathway [
17]; additionally, a high-glucose environment could also activate mast cells and induce the release of pro-inflammatory mediators, resulting in exacerbation of the pathological remodeling in DCM [
18].
Interestingly, accumulating evidence has revealed that there is a potential link between immunity and mitochondrial metabolism, and the metabolic state can affect the development of inflammation through changing the immune microenvironment [
19]. Typical T cell activation is accompanied by the up-regulation of insulin receptors and glycolytic enzymes [
20].High levels of insulin can impair the function of regulatory T cells (Tregs) and inhibit their suppressive function towards inflammatory response via regulating the AKT/mTOR signaling pathway [
21]. Both mitochondrial metabolism and immune-inflammation are key pathogeneses of DCM, but their crosstalk in DCM have not yet been reported and require further exploration.
Bioinformatics allows for screening of molecules which show a difference between patients and healthy individuals from microarray data that vary at multiple levels. It is appreciated as an effective research method for exploring the potential molecular mechanism of disease. With this method, the current study analyzed how mitochondria-related genes promote the development of DCM and correlate to the immune infiltration based on associated microarray data from GEO database (GSE4745, GSE5606, and GSE6880). Additionally, the relationship between hub mitochondria-related genes and immune infiltrates in DCM was investigated to help better understand the underlying immunometabolism during disease development.
Methods
Microarray data retrieval
DCM datasets were obtained from the public repository NCBI GEO (
http://www.ncbi.nlm.nih.gov/geo) [
22] using "diabetic cardiomyopathy" and "diabetic heart" as the search queries. We screened them further based on information such as sequencing type (transcriptology), animal species (Rattus norvegicus), sample source (ventricle), and modeling time. Finally, GSE4745, GSE5606 and GSE6880 were obtained. The GSE4745 ([RG_U34A] Affymetrix Rat Genome U34 Array) is generated by the GPL85 platform that contains 24 left ventricular (LV) samples from rattus norvegicus. To better analyze the differential genes between DCM group and control (CON) group, 8 samples collected on day 42, including 4 DCM samples and 4 CON samples, were selected for analysis [
23]. The GSE5606 ([Rat230_2] Affymetrix Rat Genome 230 2.0 Array) is generated by the GPL1355 platform and composed of 14 LV samples from DCM rats (n = 7) and CON rats (n = 7) [
24]. The GSE6880 ([RAE230A] Affymetrix Rat Expression 230A Array) is generated by the GPL341 platform comprising 6 LV samples from DCM rats (n = 3) and CON rats (n = 3) [
25].
Acquisition of microarray data and identification of differentially expressed genes (DEGs)
Data of each microarray were accessed from GEO using R package "GEO query". DEGs from each microarray were obtained with R package "limma" as implemented by GEO2R online tool (
https://www.ncbi.nlm.nih.gov/geo/geo2r/) [
26], and all identified DEGs met p < 0.05 and |log2 (Fold-change)|≥ 1. Resulting DEGs were visualized by Volcano Plot using R package "ggplot2" [
27] and Heatmap using R package "ComplexHeatmap" [
28].
Functional enrichment analysis
Gene Set Enrichment Analysis (GSEA) [
29] was applied using R package "clusterProfiler" [
30], with the "c2.cp.v7.2.symbols.gmt" (
https://www.gsea-msigdb.org/gsea/msigdb/index.jsp) as the reference gene set, the number of permutations as 10,000, and the threshold of significance as 10. The results were visualized with R package "ggplot2" [
27].
Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were accomplished in DEGs with R package "clusterProfiler" [
30], and the items with P < 0.05 in Benjamini–Hochberg test were regarded has having statistical significance. The results were visualized by Chordal and Circle graphs using R packages "ggplot2" [
27] and "GOplot" [
31].
Identification of mitochondria-related DEGs (MitoDEGs)
The mitochondrial protein database, MitoCarta3.0 (
http://www.broadinstitute.org/mitocarta) [
32], was visited to obtain 1,140 mitochondria-localized genes. MitoDEGs were obtained via intersecting the DEGs from each microarray and the 1,140 mitochondria-localized genes using a Venn Diagram, and they were visualized as a Heatmap with R package "ggplot2" [
27]. The overlapped MitoDEGs among the three microarrays were eventually obtained.
Analysis of protein–protein interactions (PPI) and identification of Hub genes
The overlapped MitoDEGs were processed for PPI analysis with STRING database (
https://string-db.org/) [
33], and the resulting interactions were visualized as a network using Cytoscape 3.8.2 [
34]. Hub MitoDEGs were screened out using the plug-ins CytoHubba and MCODE as implemented by the Cytoscape 3.8.2.
Acquisition of genes potentially key to DCM and HF
The CTD database (
http://ctdbase.org/) [
35] assembles interaction data between chemicals, gene products, functional phenotypes, and diseases, affording great convenience to research into disease-associated environmental exposures and potential mechanisms of action of drugs. With the CTD data, the link between hub MitoDEGs and the risk of developing DCM or HF was analyzed.
Prediction of a hub MitoDEGs-Transcription factors (TF) -miRNAs network
To explore the upstream regulators of hub MitoDEGs, TFs of hub MitoDEGs were predicted with the plug-in iRegulon of the Cytoscape 3.8.2 [
36], and miRNAs of hub MitoDEGs were predicted using the miRWalk database (
http://mirwalk.umm.uni-heidelberg.de) [
37]. The hub MitoDEGs, resulting TFs and miRNAs were visualized as a network by the Cytoscape 3.8.2.
Immune infiltration analysis
The gene matrices of GSE5606 and GSE6880 original datasets were combined using the Perl script and normalized after elimination of the batch effect and the heterogeneity induced by different platforms with R package "sva" [
38]. The normalized gene expression matrix was used for further immune infiltration analysis. The GSE4745 dataset was excluded from the analysis, as the number of genes in GSE4745 was significantly less than that in the other two datasets, which might result in bias results.
The ImmuCellAI (
http://bioinfo.life.hust.edu.cn/web/ImmuCellAI) estimates the infiltration abundance of 36 immune cell types based on RNA-Seq data or gene-expression profiles from microarray data [
39].The normalized gene expression matrix was uploaded to the ImmuCellAI for analysis of immune infiltration, with Wilcoxon rank sum test used for between-group comparisons. Spearman correlation analysis was applied to explore the link between MitoDEGs/hub MitoDEGs and the immune cells.
Construction of animal models with DCM
The animal procedure was performed in strict accordance with
The Guide for Care and Use of Laboratory Animals [
40] and with the approval from the Laboratory Animal Ethics Committee. Ten Sprague–Dawley male rats, weighing 200 ± 20 g, were housed in the laboratory animal center of our hospital. The Sprague–Dawley rat model of type 2 DM was generated using a high-fat diet combined with a low-dose STZ injection [
41‐
43]. The rats were allowed to acclimate for 1 week with free access to diet and water ad libitum in an environment that provided a relative temperature of 24 ℃, a relative humidity of 50–60%, and a 12 h/12 h light/dark cycle. Following that, the rats were divided into the CON and DCM groups by random assignment. The total modeling time was 16 weeks. Rats in the CON group were fed normal diet, while rats in the DCM group were fed high-fat diet containing 60 kcal% fat, 20 kcal% protein, and 20 kcal% carbohydrate. After 4 weeks, streptozotocin (STZ) (40 mg/kg, Solarbio) was intraperitoneally injected in rats of the DCM group to induce T2DM, and citric acid buffer at the same dose was administrated in rats of the CON group. One week after injection, blood glucose was measured from the tail vein, and a random glucose level > 16.7 mmol/L was indicative of successful modeling. Tissue samples were obtained after another 12 weeks of feeding. Blood glucose and body weight were monitored during modeling, and echocardiography and measurement of tibia length were performed before sampling.
Echocardiography
Rats were anesthesized with intraperitoneal injection of 30 mg/kg pentobarbital. Echocardiography was performed with the transducer of a high-resolution imaging system. LV parameters, including left ventricular ejection fraction (LVEF), fraction shortening (FS), left ventricular internal diameters at systole (LVIDs) and diastole (LVIDd), were measured from long- /short-axis images of the LV. Cardiac function was assessed by analysis of data of 3–5 cardiac cycles.
RNA extraction and qRT-PCR
Total RNA was extracted from cardiac tissue using Trizol and reversely transcribed into cDNA using a reverse transcription kit (Roche). qRT-PCR was fulfilled with a SYBR Green (Roche). The primers used for amplification were shown in Table S1. Target gene expression relative to GAPDH gene was shown as 2ΔΔCt.
Western blotting
Myocardial tissue was used to extract the samples, which were then boiled in a loading buffer for five minutes. 10% SDS-PAGE was used to separate the proteins. Primary antibodies were grown on the PVDF membranes overnight at 4 °C. The primary antibodies used in this study were Actin (Dilution 1:1000, M20011, Abmart), Pdk4 (Dilution 1:1000, YN5701, Immunoway), Hmgcs2 (Dilution 1:5000, ab137043, abcam), Decr1 (Dilution 1:1000, A13014, ABclonal), and Ivd (Dilution 1:2000, 10822-1-AP, ProteinTech). The membranes were incubated with the secondary antibody for 1 h at room temperature. Finally, the membranes were detected by the ECL system.
Immunohistochemistry
Dewaxed, rehydrated, and antigen retrieval were performed on paraffin slices. Methanol was used to inactivate endogenous peroxidase for 15 min. The slices were then sealed after being treated with 5% BSA for 1 h, followed by an overnight incubation with primary antibodies. The primary antibodies used in this study were Pdk4 (Dilution 1:50, YN5701, Immunoway), Hmgcs2 (Dilution 1:200, ab137043, abcam) Decr1 (Dilution 1:50, A13014, ABclonal) and Ivd (Dilution 1:500, 10822-1-AP, ProteinTech). The paraffin slices were cultured with secondary antibodies the next day for one hour at 37 °C. Finally, the slides were stained with a DAB Detection Kit, and the counterstained with haematoxylin.
Correlation between hub MitoDEGs and cardiac function
Correlation between hub MitoDEGs and LV parameters (EF%, FS%, and LVIDs) was analyzed using the Pearson algorithm, and the results were visualized with R package "ggplot2"[
27].
Statistical analysis
Data are presented as the mean ± standard deviation (SD) of four independent experiments, and were analyzed using GraphPad Prism 8.0 (GraphPad Inc, San Diego, USA). The Shapiro–Wilk test was used to check data normality. The student's t-test was used to evaluate the difference between the two groups when the data were in accordance with the normal distribution. P value < 0.05 was considered to be statistically significant.
Discussion
The number of DM patients has grown worldwide at an alarming rate. DM commonly occurs with target organ damage that leads to a poor prognosis, and it is tightly linked to the initiation and development of HF [
44]. It has been proven that the risk of developing HF in DM patients is associated with the presence of DCM [
45]. However, it remains elusive about the pathogenesis of DCM, and there is a paucity of effective therapeutic strategies. In this context, strengthening our understanding on DCM pathogenesis and looking for potential therapeutic targets are in urgent need. With multiple bioinformatics methods, the present study firstly obtained DEGs from the three DCM-related microarray datasets from GEO and found that the DEGs were enriched in pathways associated with mitochondrial metabolism, immune-inflammation, and collagen synthesis. Mitochondrial dysfunction and metabolic abnormality have been proven to play a role in cardiac hypertrophy and myocardial fibrosis [
46]. In addition, various activities of immune cells, such as transition from macrophages to fibroblast-like cells [
47], B-cell infiltration [
48], and transition between T lymphocyte subsets (Th17 to Treg) [
49], are also critical for pathogenesis of myocardial fibrosis. Based on the findings, our study aimed at analyzing the regulatory roles of mitochondrial metabolism and immune dysregulation in the occurrence and development of DCM and exploring related targets. The findings of the study may help us better understand the mitochondrial metabolism, immunity, and their crosstalk in DCM.
Presently, mitochondria-related genes in DCM have not yet been reported by bioinformatics studies. For the first time, our study applied the MitoCarta 3.0, an authoritative database of mitochondrial proteome, to obtain mitochondria-related genes, and then identified 9 hub MitoDEGs with had a strong correlation with DCM or HF. To validate our findings, DCM rats were modeled. Expression analysis revealed four genes, including Pdk4, Hmgcs2, Decr1, and Ivd, which showed a consistent expression trend as that detected by prior bioinformatics analysis. Additionally, we found that the up-regulation of Pdk4, Hmgcs2, Decr1 and the down-regulation of Ivd were significantly associated with the reduction in cardiac function.
Mitochondrial metabolic disorder is one of the important pathogeneses of DCM [
44], while Pdk4, Hmgcs2, Decr1, and Ivd are enzymes essential for mitochondrial metabolism. In DCM, the most significant metabolic disorders in myocardial tissues are decreased glucose utilization and increased fatty acid oxidation, which can lead to cardiac lipotoxicity, myocardial fibrosis, and effects on cardiac function. Pdk4 (Pyruvate dehydrogenase kinase 4) is localized to the mitochondrial matrix and participates in fatty acid oxidation as a key enzyme [
50]. Studies found that Pdk4 showed increased expression in myocardial tissues of DM mice [
51], and it could be used as a therapeutic target for DM [
52,
53] due to its role as a key target genes of the PPARα signaling pathway [
54,
55]. In addition, specific expression of Pdk4 could induce insulin resistance, reduction in myocardial glucose oxidation and increase in fatty acid oxidation [
56,
57]. To the contrary, suppression of Pdk4 activity could lead to reduced mitochondria-associated ER membranes (MAM) formation and improve insulin signal transduction through preventing the MAM-induced mitochondrial Ca2 + accumulation [
58]. Other than the role in mediating metabolic reprogramming, Pdk4 also has implications for cell respiration by playing a role in regulation of mitochondrial dynamics [
59]. Hmgcs2 (3-hydroxy-3-methylglutaryl-CoA synthase 2) is also distributed to the mitochondrial matrix and acts as a rate-limiting enzyme in ketogenesis [
60]. George A.Cook et al. [
61] found that Hmgcs2 was increasingly expressed in DCM rats, consistent with the present study. Another study noted significantly increased expression of Hmgcs2 enzyme in the right ventricle in cases of arrhythmogenic cardiomyopathy, suggesting enhanced ketoacid metabolism, and it also reported concurrent elevation of plasm β-hydroxybutyrate. The results indicated that up-regulation of Hmgcs2 enzyme was predictive of occurrence of major adverse cardiovascular events and disease progression [
62]. However, there was a study which demonstrated reduced cardiac content of Hmgcs2 in non-diabetic patients with end-stage HF [
63]. We speculated that the discrepancy might be due to the difference in cardiac metabolic substrates between diabetic and non-diabetic cases [
64]. Decr1 (2,4-dienoyl-CoA reductase 1) is a mitochondrial enzyme involved in degradation of poly-unsaturated fatty acids [
65]. Most of the existing studies concentrated on its role in lipid metabolism in tumor cells [
65‐
67], while only a few was performed in non-diabetic HF [
68,
69]. Therefore, further research is in demand to explore the role of Decr1 in DCM. Ivd (Isovaleryl-CoA dehydrogenase) is another mitochondrial enzyme with implications for metabolism of the branched chain amino acids leucine [
70]. Previous research revealed that leucine-enriched diet was conducive to improving the cardiac injury and dysfunction caused by cancer cachexia [
71] and anti-tumor drugs [
72]. Furthermore, circulating levels of branched chain amino acids were proven as independently associated with the incidence of HF in diabetic patients [
73].
The metabolic status and immune processes are interconnected [
74]. Immune dysregulation is common in DCM and plays a role in disease progression. In the present study, we used the ImmuCellAI algorithm to analyze immune cell infiltration and found higher enrichment of multiple dendritic cells (Dendritic cells, MoDC, cDC1, pDC, and cDC2) in the CON group than the DCM group. Dendritic cells are specialized antigen-presenting cells that serve as important mediators of immune responses [
75], and the number was reduced in both type 1 and 2 DM patients [
76,
77]. It was reported that dendritic cells were protective immunomodulators playing a role during the healing from myocardial infarction. In addition, dendritic cells tended to accumulate in infarct border zone after myocardial infarction and simultaneously mediated the regulation of homeostasis by monocytes and macrophages [
78]. In human infarcted myocardial tissues, the reduced number of dendritic cells was reported as associated with the recruitment of pro-inflammatory monocytes, increase in macrophages, impairment of reparative fibrosis, and the cardiac rupture after myocardial infarction [
79]. In all, dendritic cells protect the heart via regulating the recruitment of various types of immune cells. The current study also found that B cell, Marginal Zone B, and Memory B were highly abundant in the DCM group. B cells maintain the bridge between innate and adaptive immunity through their antigen-specific responses, and they are also conducive to sustaining the chronic inflammation in DCM [
80]. Animal experiments revealed that B cells regulated the composition of the cardiac leukocyte pool, and B cell-deficient mice had a smaller fibrotic area while a higher LVEF [
81]. Another study found that B cell depletion was accompanied by significant reductions in TNF-α, IL-1β, IL-18, and apoptosis in myocardial cells, and further introduction of B cells worsened inflammatory response and cardiac function [
82]. Collectively, B cells are critical for the pro-inflammatory environment of the failing heart tissue and myocardial injury. There were also some studies showing that increase in neutrophil-to-lymphocyte ratio was associated with the incidence of subclinical DCM [
83]; impaired Th/Treg balance and increased ventricular infiltration of T cells exacerbated the cardiac hypertrophy and fibrosis in T2DM [
84,
85]; M1 macrophages potentiated DCM progression via secreting inflammatory factors to induce insulin resistance [
86].
Mitochondrial metabolism can have a huge impact on the fate and function of immune cells. Correlation analysis of the study indicated that Pdk4, Hmgcs2, and Decr1 were positively associated Marginal Zone B while negatively associated with dendritic cells. In addition, Ivd was positively associated with CD8 Tem. This is consistent with our findings that dendritic cells had a lower enrichment in the DCM group than the CON group and had significant enrichment in B cells. The findings of the study deepen our understanding about the link between mitochondrial metabolism and immune cells in DCM.
In this study, the interaction between mitochondrial metabolism and the immune microenvironment was found for the first time through bioinformatics analysis of DCM. Screening and verification of Pdk4, Hmgcs2, Decr1 and Ivd provide potential molecular targets for deep exploration of immunometabolism in DCM. There are some limitations to our study. Firstly, we have only validated the hub genes in the DCM rats and lack the support of clinical data. Secondly, although a rigorous bioinformatics analysis was conducted in the present study, we did not conduct further experiments to verify the effects of mitochondrial metabolism genes on the immune microenvironment and cardiac function. Hence, the specific mechanism of immunometabolism regulation in DCM still needs to be further explored in vivo and in vitro. This novel direction will be the focus of our subsequent study.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.