Uncovering the gene regulatory network of type 2 diabetes through multi-omic data integration

Type 2 diabetes (T2D) is a chronic metabolic disease distinguished by insulin resistance and elevated blood glucose levels. As a global endemic, recent data from the Centers for Disease Control and Prevention (CDC) [1] suggested that as of 2019, roughly 28.7 million people in the United States (8.7% of the total U.S. population) were diagnosed with diabetes, of which about 90–95% have T2D. As prolonged hyperglycemia is a high-risk factor for heart disease, chronic kidney disease (CKD), and nerve damage, T2D imposes a substantial economic burden on society [2]. In addition, calculations based on epidemiological data suggested that the expenditure on diabetes in the U.S. was approximately $327 billion in 2017, including $237 billion in direct medical costs and $90 billion in lost productivity [3].

Due to the complexity of the onset and progression of T2D and the tandem with various diseases, analysis from multiple levels could exponentially augment our understanding of its pathophysiological mechanism. Including genomics, epigenomics, transcriptomics, etc., multi-omics analysis [4] can provide a list of disease-related differences that can be used as biomarkers of the disease process and uncover critical pathways in disease. For instance, Yang-Tay et al. discovered the DNMT1-NT5C2-insulin receptor pathway using the DNA methylation array data, demonstrating that DNMT1 is relevant to the susceptibility of T2D patients [5]. Besides, through the single-cell transcriptome analysis, Lawlor et al. revealed that PP/gamma cells in T2D patients could integrate central and peripheral hunger and satiety signals, which increases our accurate knowledge of the molecular components of rare islet cells [6]. However, unlike single-omics studies, multi-omics studies can be more comprehensive and accurate. At present, multi-omics studies on T2D have received extensive attention. For example, to better understand preT2D status, Wenyu et al. [7] conducted a cohort study of 106 normal individuals and individuals with prediabetes. Through conjoined analysis of transcriptomic and proteomic data, etc., the study indicated a new sight in that the gut microbiota of insulin-resistant individuals may have a decreased response to respiratory viral infection. This could lead to the emergence of chronic inflammation, which promotes the progression of T2D in preT2D patients. In addition, by combining GWAS data with other multi-omics datasets, Yon Jung [8] revealed the common pathways shared by IGF-I and IR, such as glycosaminoglycan biosynthesis, etc. It provided assistance for a more comprehensive understanding of the molecular mechanism of the IGF-I/IR axis. Although previous observations are of great significance, by comprehensively using the data of genomics, transcriptomics, and epigenomics to explore disease-related gene expression signatures, our cognition of T2D and the clinical transformation of drug discovery can both be advanced.

In this study, more accurate potential key genes and their regulatory mechanisms in T2D were identified by constructing gene regulatory network through multi-omics analysis, which contributes to demonstrating pathology, and identifying drug targets of T2D (Fig. 1).

The current paper demonstrated the molecular mechanisms associated with T2D progression by combining transcriptomic, genomic, epigenetic, and proteomic data. At the same time, some less-reported genes and pathways that may play key regulatory roles in T2D have been identified, which could provide new biomarker options for T2D research. In addition, this study conducted drug repositioning based on multi-omics data to facilitate the clinical translation of potential T2D therapeutics.

To identify markers for T2D and better understand the underlying pathways, we performed MSEA using multi-omics data. In the results, GO terms were enriched in “response to hypoxia” and “response to decreased oxygen levels”, implying the stress of oxygen deprivation. Hypoxia could activate the hypoxia-inducible factor rapidly, leading to the conversion of glucose utilization from aerobic to anaerobic metabolism, resulting in β-cell dysfunction [21, 22]. Besides, Li et al. found that hypoxia may cause intensive apoptotic injury of β cells by destroying islet vascular integrity [23]. Related to the above, “gap junction assembly” and “mitochondrial outer membrane” were also enriched. β cells are connected by gap junction, which provides electric coupling between β cells, thus promoting the regulation of electrical activity and insulin secretion [24]. Studies have confirmed that related to the abnormal expression of gap junction proteins, the breakdown of mitochondrial redox balance in β cells under long-term hyperglycemia will accelerate the dysfunction of β cells [25]. On the other hand, “sulfur compound binding” was also enriched in the GO-based list. Sulfur-containing compounds include sulfur-containing amino acids, glutathione, etc. Elevated cysteine of sulfur-containing amino acids is associated with a doubling of the risk of insulin resistance [26]. In addition, data from a prospective cohort study showed that diabetic patients had higher cystathionine and plasma total cysteine and lower antioxidants such as taurine [27]. These findings suggest that sulfur-containing amino acids may interfere with blood glucose levels through oxidative stress. Moreover, the pathways “Leukocyte Transendothelial Migration” and “Reactome Signaling By Interleukins” were also significantly enriched. For islet vascular damage in diabetes, neutrophil transmigration across TNF-activated endothelial monolayers can be accelerated by co-clustering L-selectin with PECAM-1. And another crucial step is the shedding of L-selectin activated by Akt family kinases, p38 MAPK signaling pathways, etc. [28]. As previously mentioned, increased secretion and chemotaxis of neutrophils would stimulate β cell apoptosis by reducing insulin signaling transduction, promoting the occurrence of ROS-NLRP3 inflammasome-IL-1β [29].

Elucidation of T2D driving molecular profiles through integrative multi-omic analysis, including genomic, epigenomic, and transcriptomic analysis, was the primary focus of this study. The most prominent finding to emerge from the research is that putative ten hub genes are associated with T2D. Among these, COL5A1, IRF7, CD74, and HLA-DRB1 expression was suggested to have diagnostic value in T2D, and the expression levels of PSMB9, COL1A1, and COL4A1 were significantly higher in T2D after validation. Besides, the enrichment analysis showed the hub genes were significantly associated with immune-related terms and T2D-related terms, which corroborates the findings of a great deal of the previous work [30‐32]. Chemokines and cytokines are jointly involved in the occurrence and development of T2D [33, 34], such as serum TNF-α, adiponectin, Growth factor 19/21, Interleukin-1 beta (IL-1β), Interleukin-6 (IL-6), Interleukin-18 (IL-18), and C-reactive protein (CRP), which play a central role in the development of T2D. According to KEGG analysis, the AGE-RAGE signaling pathway plays a vital role in T2D. In T2D patients, elevated blood glucose leads to advanced glycation end products [35]; the products of nonenzymatic glycation/oxidation of proteins/lipids are signal transduction ligands for Receptor for AGE (RAGE), which accumulate in the vessel wall. Furthermore, the recruitment of inflammatory cells bearing Calgranulin B (S100A9), also ligands for RAGE, augments vascular dysfunction and can subsequently exacerbate the progression of T2D [36, 37]. In addition, we found that 8 TFs may regulate the expression of these genes. We further verified that NFKB1 is highly expressed in T2D patients, which coordinately participated in regulating four hub genes (IRF7, PSMB9, CD74, and COL1A1).

In accordance with the present results, previous studies have demonstrated that HLA-DQB1 [38], COL1A1 [39], COL3A1 [40], COL4A1 [41], CD74 [42], and HLA-DQA1 [38] are highly related to T2D pathogenesis. While PSMB9, IRF7, and COL5A1 have not been previously reported to be associated with T2D in the literature. PSMB9 is a multicatalytic proteinase complex with a highly ordered ring-shaped 20S core structure, which encodes a member of the proteasome B-type family, also known as the T1B family. An essential function of a modified proteasome, the immunoproteasome, is the processing of class I MHC peptides, which have critical roles in the immune response. The activation of PSMB9 may be related to the expression of CTCF, which is shown in the results. CTCF, also called CCCTC-binding factor, encodes a transcriptional regulatory protein with 11 highly conserved zinc finger (ZF) domains. It has been revealed that CTCF-mediated chromatin accessibility changes could help to increase the transcription of genes related to important functions of pancreatic β cells, thereby increasing insulin secretion and improving T2D [43]. Diseases associated with the PSMB9 gene in GWAS datasets from the DISEASES Experimental Gene-Disease Association Evidence Scores dataset revealed a potential association between PSMB9 and type 1 diabetes mellitus with a standardized value of 1.20753. Besides, Phenotypes associated with the PSMB9 gene by text-mining GWAS publications from the HuGE Navigator Gene-Phenotype Associations dataset also suggested the potential association between PSMB9 mutations and type 1 diabetes. IRF7 encodes interferon regulatory factor 7, a member of the interferon regulatory factor (IRF) family, which has been shown to play a role in the transcriptional activation of virus-inducible cellular genes, including interferon β chain genes. In the results of DNase-seq, the cis-acting elements located near the TSS region of IRF7 showed high chromatin accessibility. It means that the transcription of IRF7 may be regulated in many ways, such as H3K4me3, H3K27ac, and H3K9ac, which promote the activation of transcription. A previous study has reported that highly expressed IRF7 in the pancreas is implicated in immunoinflammatory diseases such as autoimmune pancreatitis [44] and pancreatic ductal adenocarcinoma [45]. In addition, a study by Hemin et al. demonstrated that STAT1-IRF7-MHC I complex axis was crucial for IFN-α signaling in islets and created positive feedback through IRF7-STAT2 cascade amplifying signals which accelerated the process of type 1 diabetes. COL5A1 encodes an alpha chain for one of the low abundance fibrillar collagens. Previous work revealed that Col5A1, Nqo1, and Notch2 modulated by Ast may promote insulin-releasing balance, relieve insulin resistance, and maintain normal size in marginal-zone B cells [46]. Besides, diseases associated with COL5A1 gene/protein from the curated CTD Gene-Disease Associations dataset suggested the potential association between COL5A1 and diabetes mellitus with a standardized value of 1.24401 [47]. Hence, the above results provide meaningful clues for the further study of PSMB9, IRF7, and COL5A1 in the occurrence and development of T2D, and further exploration is needed.

Due to the lack of effective/safe and less expensive drugs, drug repositioning appears to be the best tool for finding proper targets and predicting latent drugs in the therapy for T2D and related complications. Finally, 17 compounds were expected to have potential therapeutic effects on T2D. Among these, sunitinib [48] is a multi-target receptor tyrosine kinase inhibitor of vascular endothelial growth factor receptor (VEGFR), platelet derived growth factor receptor (PDGFR), etc. It is often used as a chemotherapeutic agent to treat renal cell carcinoma (RCC) and pancreatic neuroendocrine tumors. Considering that VEGF/VEGFR, PDGF/PDGFR related signal pathways play essential roles during the development of T2D, such as insulin resistance [49], the expression of iron metabolism related-proteins [50], islet cell inflammation [51], further studies to investigate potential therapeutic benefits for sunitinib in T2D are warranted.

Due to the heterogeneity of T2D [52], future T2D treatment urgently needs to be personalized. Nowadays, nanotechnology has shown great prospects in T2D pharmacological intervention, such as extending the release of anti-diabetic peptides through the hydrogel system [53], oracle delivery of nuclear acid therapy with higher stability [54], etc. In addition, the use of versatile drug delivery nanocarriers after multi-level specific biomarker recognition based on multi-omics data may increase the targeting effect of drugs and promote the realization of personalized T2D treatment [55].

Despite these promising results, questions remain. The results of RT-qPCR on the predicted hub genes could only show that the expressions of CD74 and PSMB9 were up-regulated in the T2D group, while the expressions of other genes were not detected (Additional files 3, 4). In addition to the poor effect of RNA extraction, since the samples used for RNA extraction were human serum rather than islets, the greater heterogeneity of the gene expression levels of the two could be the cause. For that, it would be problematic to demonstrate the expression of predicted genes in islets due to the inaccessibility of normal human islets and the potential ethical issues involved.