Molecular Medicine Reports Molecular Medicine Reports 1791-2997 1791-3004 D.A. Spandidos 10.3892/mmr.2018.8804 mmr-17-06-7636 Articles Integrated analysis of the gene expression profile and DNA methylation profile of obese patients with type 2 diabetes ShenJuan 1 * ZhuBin 2 * Nursing Department, Suzhou Vocational Health College, Suzhou, Jiangsu 215000, P.R. China Anesthesiology Department, The People's Liberation Army 100 Hospital, Suzhou, Jiangsu 215000, P.R. China Correspondence to: Dr Juan Shen, Nursing Department, Suzhou Vocational Health College, 28 Kehua Road, Suzhou, Jiangsu 215000, P.R. China, E-mail: jshen@szhct.edu.cn

Contributed equally

 062018 28032018 17 6 7636 7644 05022017 20112017 Copyright: © Shen et al. 2018 This is an open access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made.

In order to better understand the etiology of obese type 2 diabetes (T2D) at the molecular level, the present study investigated the gene expression and DNA methylation profiles associated with T2D via systemic analysis. Gene expression (GSE64998) and DNA methylation profiles (GSE65057) from liver tissues of healthy controls and obese patients with T2D were downloaded from the Gene Expression Omnibus database. Differentially-expressed genes (DEGs) and differentially-methylated genes (DMGs) were identified using the Limma package, and their overlapping genes were additionally determined. Enrichment analysis was performed using the BioCloud platform on the DEGs and the overlapping genes. Using Cytoscape software, protein-protein interaction (PPI), transcription factor target networks and microRNA (miRNA) target networks were then constructed in order to determine associated hub genes. In addition, a further GSE15653 dataset was utilized in order to validate the DEGs identified in the GSE64998 dataset analyses. A total of 251 DEGs, including 124 upregulated and 127 downregulated genes, were detected, and a total of 9,698 genes were demonstrated to be differentially methylated in obese patients with T2D compared with non-obese healthy controls. A total of 103 overlapping genes between the two datasets were revealed, including 47 upregulated genes and 56 downregulated genes. The identified overlapping genes were revealed to be strongly associated with fatty acid and glucose metabolic pathways, in addition to oxidation/reduction. The overlapping genes cyclin D1 (CCND1), PPARG coactivator α (PPARGC1A), fatty acid synthase (FASN), glucokinase (GCK), steraroyl-coA desaturase (SCD) and tyrosine aminotransferase (TAT) had higher degrees in the PPI, transcription target networks and miRNA target networks. In addition, among the 251 DEGs, a total of 35 DEGs were validated to be being shared genes between the datasets, which included a number of key genes in the PPI network, including CCND1, FASN and TAT. Abnormal gene expression and DNA methylation patterns that were implicated in fatty acid and glucose metabolic pathways and oxidation/reduction reactions were detected in obese patients with T2D. Furthermore, the CCND1, PPARGC1A, FANS, GCK, SCD and TAT genes may serve a role in the development of obesity-associated T2D.

 DNA methylation profile gene expression profile obesity-associated type 2 diabetes protein-protein interaction network regulatory network Introduction

Type 2 diabetes (T2D), characterized by an inadequate β-cell response to progressive insulin resistance, is a highly prevalent disease affecting ~9% of the global population, and is fast-becoming a worldwide epidemic (1,2). The clinical symptoms of T2D include hyperglycemia, obesity, hypertension and hyperlipidemia. Furthermore, T2D may induce disease-specific complications, including blindness, renal failure and increased risk of cardiovascular disease, which may result in a reduced quality of life and an increased mortality rate of patients with T2D (3,4).

T2D is a complex disease that may be attributed to the interplay between environmental and genetic risk factors (5). Poor diets and sedentary lifestyles are prominent environmental contributors leading to the development of T2D (6). Epigenetic factors have been revealed to be heavily implicated in the complex interplay between environmental signals and intrinsic genetic alterations (7). DNA methylation is an epigenetic modification most commonly associated with cysteine-phosphate-guanine (CpG) sites situated within the promoter region, and degrees of organismal DNA methylation are changeable depending on environmental factors. Furthermore, DNA methylation may modulate gene expression without altering the sequence of DNA via suppression of DNA transcription or modification of the surrounding chromatin. Methylation may suppress transcription by modulating the binding of transcription factors (TFs) to DNA, and via recruitment of methyl binding proteins and transcriptional corepressors (8). Therefore, DNA methylation modification represents a link between environmental risk factors and disease progression by influencing gene transcription patterns and, subsequently, organ function. Typically, advancing age, physical inactivity, weight gain and obesity are primary risk factors for the development of T2D (9). In addition, patients suffering from metabolic syndromes with inherent symptoms of glucose intolerance, insulin resistance and abdominal obesity are considered to be in prediabetic state, which may ultimately develop into T2D (10). Previous studies have revealed an association between DNA methylation patterns and alterations in body weight and physical activity. Furthermore, CpG markers of DNA methylation are biomarkers for metabolic syndrome (11–13). Alterations in metabolite levels, including choline, betaine and methionine, are implicated in methylation pathways in the liver (14,15), and choline-associated metabolites have been demonstrated to be implicated in the pathological development of T2D (16). Therefore, DNA methylation has been hypothesized to have an involvement in the pathogenesis of T2D. Furthermore, the negative correlation between increased methylation levels of β-cell specific genes, including pancreatic and duodenal homeobox1 and insulin, and the expression levels of their corresponding proteins, have previously been detected in the pancreatic islets of patients with T2D (17,18). Therefore, there is an incentive to investigate the potential implications of DNA methylation and the associated gene expression pattern modifications with regards to the pathogenic onset of T2D.

T2D is a highly complex multisystem disease. Reduced rates of muscular glycogen synthesis in patients with insulin-dependent diabetes may be induced by defective glucose transport/phosphorylation (19). Furthermore, alterations in mitochondrial gene transcription patterns in skeletal muscle are closely associated with insulin-dependent T2D (20). In addition, the liver is implicated in the regulation of lipid and glucose metabolism, disorders of which frequently occur in non-alcoholic fatty liver disease (NAFLD) and T2D (21). In the present study, a systematic analysis was performed using publicly-available online genome-wide methylome and transcriptome data from liver tissues from age-matched healthy and obese T2D men, uploaded by Kirchner et al (22), in order to identify disease-associated genes and to better understand T2D at the molecular level. Unlike the study by Kirchner et al (22), the present study aimed to reveal the protein-protein interaction (PPI), TF target and microRNA (miRNA) target networks among the differentially-expressed genes (DEGs), in order to develop a more comprehensive understanding of protein function associated with T2D.

 Materials and methods <sec> <title>Microarray data

The raw data on gene expression were downloaded from the Gene Expression Omnibus (GEO) database (www.ncbi.nlm.nih.gov/geo), accession no. GSE64998. This dataset, including 21 samples (liver biopsies from six non-obese, eight obese non-diabetic and seven obese T2D men), were collected based on the GPL11532 platform (HuGene-1_1-st) Affymetrix Human Gene 1.1 ST Array [transcript (gene) version]. The data on liver tissues isolated from six non-obese men and seven obese T2D men were extracted. These data were uploaded by Kirchner et al (22); their study was conducted according to the principles described in the Declaration of Helsinki, the regional ethics committee at the Karolinska Insitute (Solna, Sweden) approved the study, and all participants provided informed written consent.

Furthermore, the methylation profile data based on the GPL13534 platform [Illumina Human Methylation 450 BeadChip (HumanMethylation450_15017482)] were downloaded from the GEO database, accession no. GSE65057. The data on liver tissues isolated from seven non-obese controls and nine obese T2D samples were extracted from GSE65057, which were uploaded by Kirchner et al (22).

 Data preprocessing

Raw expression profile data in the CEL format were preprocessed using the Oligo package in R (23), which included format transition, missing value interpolation, background correction and data quantile normalization.

The RnBeads Package (24), which used β-values in order to characterize the degree of DNA methylation, was applied for analysis of the downloaded methylation microarray data. Initially, the Infinium probes were manipulated using the Methylumi package (25). Following this, background correction was performed with the normal-exponential convolution using the out-of-band probes method (26), and normalization was performed using the Beta MIxture Quantile dilation method (27). Finally, the probes with a detection value of P>0.01 or bead count <3, located on sex chromosomes or in regions enriched with single nucleotide polymorphisms, were deleted (28).

 DEG and differentially-methylated gene (DMG) screening

The empirical Bayes approach in the Limma package (29) was used in order to identify differing levels of DEGs and DMGs between the healthy controls and the obese T2D samples. DEGs were classified as those meeting the criteria of P<0.05 and |log fold change|≥0.5, and DMGs were those with P<0.05. Furthermore, DMGs were mapped to the DEGs in order to identify any overlaps.

 Functional and pathway enrichment analyses

The Multifaceted Analysis Tool for Human Transcriptome (www.biocloudservice.com) online tool in BioCloud, a platform storing vast amounts of bioinformatics data and providing analysis software applications, was used to perform Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway and Gene Ontology (GO) functional enrichment analyses for the upregulated genes, the downregulated genes and the overlapping genes. KEGG and GO are freely available for public use for the annotation of genes, gene products and gene sequences (30,31). P<0.05 was set as the threshold criterion.

 Construction of PPI network

The PPIs among the DEGs were analyzed using the STRING database (32) and the default parameters, and the combined score >0.4 was set as the threshold. Following this, Cytoscape software v3.2.0 (33) was used in order to visualize the PPI network, and connectivity degree analysis was performed in order to screen for hub genes (34). Furthermore, the sub-network involving overlapping genes was extracted from the PPI network.

 Construction of the regulatory network for overlapping genes involved in the sub-network

Using the iRegulon plugin (35) in Cytoscape software (33), the TFs targeting the overlapping genes involved in the sub-network were predicted. TF target pairs with a normalized enrichment score >4, calculated with iRegulon plugin, were selected, and the TF target regulatory network was constructed using Cytoscape software (33). In addition, miRNAs targeting the overlapping genes involved in the sub-network were predicted using the WebGestalt tool (36,37). Following the determination of the miRNA target pairs, the miRNA target regulatory network was visualized using Cytoscape software (33).

 Data validation of DEGs

A further GSE15653 dataset was used for validation of the DEGs already identified. The GSE15653 dataset was downloaded from GEO, and included four liver tissue samples from obese T2D patients and five healthy controls. Following this, the raw expression profile data in the CEL format from the GSE15653 dataset were preprocessed using the same methods above, and the DEGs in the obese T2D samples were also identified also using the same methods and threshold value used above. Subsequently, the shared upregulated and downregulated DEGs in the GSE15653 and GSE64998 datasets were obtained via Venn analysis, and these identified DEGs were considered to validate the genes identified from GSE64998 dataset.

 Results <sec> <title>Identification of DEGs and DMGs

Following analysis of the gene expression profiles of the non-obese and the obese T2D samples, 251 DEGs were detected, including 124 upregulated genes and 127 downregulated genes. Furthermore, 9,698 DMGs (6,021 upregulated genes and 3,677 downregulated genes) were identified in obese diabetic individuals compared with non-obese controls (P<0.05). Following the mapping of the DEGs to the DMGs, a total of 103 overlapping genes were revealed (47 upregulated genes and 56 downregulated genes) in the gene expression profiles.

 Functional and pathway enrichment analyses

In order to examine the biological functions of abnormal genes in obesity-associated T2D, GO and KEGG enrichment analyses were performed using the previously identified DEGs and the overlapping genes. Fig. 1A and B present the enriched KEGG pathways for the downregulated genes and upregulated genes, respectively; and Fig. 1C and D present the top five prevalent GO terms for the downregulated genes and upregulated genes, respectively. The upregulated genes were most significantly enriched in ‘biosynthesis of unsaturated fatty acids’ (KEGG pathway; P=9.46×10−4), and the downregulated genes were significantly enriched in KEGG pathways including ‘nitrogen metabolism’ (P=3.52×10−4) and ‘cysteine and methionine metabolism’ (P=2.89×10−2). Furthermore, the 103 overlapping genes were significantly enriched in ‘arginine and proline metabolism’ (KEGG pathway; P=5.53×10−3) and ‘oxidation-reduction reactivity’ (GO term; P=3.52×10−4; Table I).

 Construction of PPI network

PPIs were determined by STRING database analysis, and a PPI network with 116 nodes and 189 edges was generated for the DEGs (Fig. 2). The top 20 nodes with the highest degrees are detailed in Table II. Notably, fatty acid synthase (FASN), cyclin D1 (CCND1), glucokinase (GCK), stearoyl-CoA desaturase (SCD), and PPARG coactivator α (PPARGC1A) were all overlapping genes in the PPI network.

Furthermore, KEGG pathway and GO functional enrichment analyses revealed that the top 20 nodes were predominantly associated with cancer pathways, including ‘p53 signaling pathway’ (KEGG pathway; P=1.36×10−3), and ‘regeneration’ (GO term, P=1.80×10−6; Table III).

 Construction of TF target network and miRNA target network

TFs and miRNAs are able to regulate gene expression via modulation of transcriptional activation and stability of mRNA, respectively (38). In the present study, the TF target and miRNA target pairs were predicted based on the overlapping genes involved in the sub-network in order to explore their potential regulatory relationships. A total of 10 TFs were predicted and the TF target network contained 49 nodes and 161 pairs (Fig. 3). The hub genes with the highest degrees are detailed in Table IV. A total of 49 miRNAs, which may be implicated in the abnormal expression of the overlapping genes, were predicted. Following this, a miRNA target network, including 87 nodes and 180 regulatory relationships, was constructed (Fig. 4). Furthermore, the hub genes with the highest degrees were screened for (Table V).

 Validation of the expression levels of DEGs

A total of 753 upregulated DEGs and 432 downregulated DEGs were identified from the GSE15653 validation dataset using the obesity-associated T2D patients and control samples according to the same method used to identify DEGs in the GSE64998 dataset. By comparing the datasets, it was revealed that among the 124 upregulated DEGs in GSE64998, 15 genes were overlapping genes (e.g., CCND1 and FASN), while among the 127 downregulated DEGs, 20 overlapping genes [e.g., tyrosine immunotransferase (TAT)] were searched. Overlapping DEGs between both datasets were considered to represent preliminary verification of said genes in the GSE64998 dataset (Table VI).

 Discussion

The high prevalence of T2D and the severity of its associated complications raise great challenges for effective disease management (39). In China, obesity is one of the principal contributory risk factors for T2D development (9). Despite several genes and their regulatory mechanisms being suggested to be implicated in the development of obesity-associated T2D, there remains a requirement for further research in order to uncover the underlying molecular mechanisms implicated in T2D pathogenesis. In the present study, systematic analysis using gene expression patterns and DNA methylation profiles of healthy controls and obese T2D patients was performed in order to reveal hub genes, which may be involved in the pathogenesis of obesity-associated T2D. By performing PPI, TF target and miRNA network analyses, the present study demonstrated that the hub nodes CCND1, PPARGC1A, ATP citrate lyase (ACLY), TAT and FASN may be implicated in the development of obesity-associated T2D.

CCND1 encodes the cyclin D1 protein, and expression of CCND1 has marked periodicity throughout the cell cycle (40). Previous studies have demonstrated an association between the dysregulation of CCND1 expression and T2D (41–43). Microarray and reverse transcription-quantitative polymerase chain reaction data have previously revealed a higher expression pattern of CCND1 in diabetic islets compared with healthy controls (43). In accordance with this, the present study predicted from the GSE64998 and GSE15653 datasets that the expression of CCND1 was upregulated in patients with T2D. In addition, the increased methylation of CCND1 was detectable in patients with T2D (44). However, it has previously been reported that there is no association between the methylation status of CCND1 and its expression (45). Further studies are required to reveal how CCND1 is implicated in T2D pathogenesis.

PPARGC1A is a transcriptional coactivator that modulates genes associated with energy metabolism (46). Numerous studies have demonstrated the link between PPARGC1A, and the development of T2D and associated insulin resistance. Decreased PPARGC1A expression has been detected in cases of insulin resistance (47–49). Furthermore, increased DNA methylation at the site of the PPARGC1A promoter has been detected in skeletal muscle tissue and in the islets of patients with T2D (50). In addition, a negative correlation between the methylation of PPARGC1A and its expression has previously been reported (51). In line with these previous findings, abnormal expression and methylation of PPARGC1A was demonstrated in liver tissues from obese T2D patients in the present study.

In the present study, it was revealed that DNA methylation corresponds with the upregulation of ACLY, FASN, SCD and GCK expression in samples from obese patients with T2D, and had high degrees in the PPI network. ACLY, FASN, GCK and SCD are all enzymes implicated in metabolic processes. ACLY is implicated in the synthesis of cytosolic acetyl-coenzyme (Co)A in numerous tissue types (52). Furthermore, Guay et al (53) demonstrated that ACLY is a fundamental regulator of glucose-induced insulin secretion. FASN, coding for fatty acid synthase, catalyzes the synthesis of palmitate from acetyl-CoA and malonyl-CoA, producing long-chain saturated free fatty acids in the presence of nicotinamide adenine dinucleotide phosphate. Genetic alterations affecting FASN activity may be significantly correlated with T2D via modification of insulin sensitivity (54). A further interpretation of the association of FASN with T2D is that an increase in fatty acids may inhibit insulin signaling and induce metabolic insulin resistance in patients with T2D (55). Furthermore, it has previously been suggested that methylation of the FASN promoter at the 611, 096, 61780, 61778 and 61774 CpG sites may be associated with the progression of NAFLD (56), which is associated with an increased risk of T2D. SCD encodes for the stearoyl-CoA desaturase enzyme that is responsible for fatty acid biosynthesis, which is implicated in lipid-induced insulin resistance, and SCD deficiency increases insulin signaling (57), Furthermore, the expression of SCD is upregulated in diabetic fatty rats (58). In addition, it has been suggested that alterations in SCD expression as a consequence of DNA promoter methylation in morbidly obese patients are associated with the serum levels of free fatty acids (59). Furthermore, the levels of mRNA encoding ACLY, FASN and SCD were markedly upregulated in the livers of Zucker fatty rats, which are commonly used as animal models for fatty liver disease, hepatic insulin resistance and obesity investigations (60). Functional enrichment analysis in the present study revealed that abnormal FASN and SCD expression levels in obesity-associated T2D were predominantly associated with fatty acid biosynthesis metabolism and oxidation reduction. GCK is responsible for the phosphorylation of glucose in order to produce glucose-6-phosphate, which is the first step in the majority of glucose metabolic pathways (61). GCK is predominantly expressed in pancreatic β cells and hepatocytes, and it is implicated the modulation of glucose homeostasis in liver, including glucose synthesis, breakdown and storage (62). It has previously been demonstrated that fluctuations in the expression levels of GCK are a risk factor for the development of T2D (63). Furthermore, elevated levels of CpG island methylation within the GCK gene have been reported in patients with T2D (64). Therefore, the upregulation of ACLY, FASN, SCD and GCK gene expression levels, accompanied by alterations in DNA methylation levels, may be implicated in the development of obesity-associated T2D via regulation of fatty acid and glucose metabolic pathways, and involvement in oxidoreductive and insulin signaling pathways.

The downregulated genes identified in patients with T2D were significantly enriched in the cysteine and methionine metabolic pathways. A previous study reported that cysteine and methionine intake has an association with T2D (65), and that levels of oxidation at cysteine and methionine residues are markedly higher in patients with diabetes compared with nondiabetic individuals (66). The present study revealed that the downregulation of cystathionine γ lyase (CTH) and TAT genes is associated with the metabolism of cysteine and methionine in patients with T2D. Furthermore, a deficiency in CTH may impair H2S biosynthesis and vessel reactivity in T2D (67). TAT is present in the liver and catalyzes the conversion of L-tyrosine into phosphorylated hydroxyphenylpyruvate, and its expression is abnormal in diabetic rats (68). Thus, we hypothesized that the downregulated expression of CTH and TAT altered the DNA methylation levels in obese patients with T2D. However, the influence of their methylation status on the development of T2D has yet to be determined. Further research is required in order to determine whether the alteration in DNA methylation levels in CTH and TAT is correlated with their downregulated expression in T2D.

However, although the present study identified genes with altered expression and DNA methylation in T2D by reanalyzing a published dataset, a number of novel genes were demonstrated to serve potential roles in T2D. These results prove beneficial to the development of a deeper understanding of obesity-associated T2D disease progression. However, the present study had certain limitations. The sample sizes analyzed in GSE64998 and GSE65057 datasets were small. In addition, the association between the expression levels of the DEGs and patterns of DNA methylation were not investigated. Future studies may investigate the correlation between the DEGs and DMGs, and analyze the methylation sites in such DEGs using large sample sizes.

In conclusion, the present study analyzed the mRNA expression and DNA methylation profiles of healthy controls and of patients with obesity-associated T2D using a computational bioinformatics approach. Genes with abnormal expression levels were screened for and the biological functions enriched by these genes were explored. In the present study, several key genes (ACLY, CCND1, PPARGC1A, FASN, GCK, SCD, CTH and TAT) were revealed to be potentially implicated in the progression of insulin resistance in obesity and T2D. However, further experimental studies are required in order to validate the implications of these genes in obesity-associated T2D.

 Acknowledgements

The present study was supported by the Overseas Study Project of Jiangsu Province for Outstanding Young and Middle-aged Teachers and Principals of Colleges and Universities (grant no. Jiangsu Teachers 2012-13).

 References DeFronzoRAPathogenesis of type 2 (non-insulin dependent) diabetes mellitus: a balanced overviewDiabetologia35389397199210.1007/BF004012081516769 Global status report on noncommunicable diseases 2014World Health Organization: 2982014 KingGLThe role of inflammatory cytokines in diabetes and its complicationsJ Periodontol798 SupplS1527S1534200810.1902/jop.2008.080246 LloydASawyerWHopkinsonPImpact of long-term complications on quality of life in patients with type 2 diabetes not using insulinValue Health4392400200110.1046/j.1524-4733.2001.45029.x11705130 Esparza-RomeroJValenciaMEUrquidez-RomeroRChaudhariLSHansonRLKnowlerWCRavussinEBennettPHSchulzLOEnvironmentally driven increases in type 2 diabetes and obesity in pima indians and non-pimas in mexico over a 15-year period: The maycoba projectDiabetes Care3820752082201510.2337/dc15-0089262464574613913 ZimmetPAlbertiKGShawJGlobal and societal implications of the diabetes epidemicNature414782787200110.1038/414782a11742409 JaenischRBirdAEpigenetic regulation of gene expression: How the genome integrates intrinsic and environmental signalsNat Genet33SupplS245S254200310.1038/ng1089 KloseRJBirdAPGenomic DNA methylation: The mark and its mediatorsTrends Biochem Sci318997200610.1016/j.tibs.2005.12.00816403636 WangCLiJXueHLiYHuangJMaiJChenJCaoJWuXGuoDType 2 diabetes mellitus incidence in Chinese: Contributions of overweight and obesityDiabetes Res Clin Pract107424432201510.1016/j.diabres.2014.09.05925649908 DiamantMTushuizenMEThe metabolic syndrome and endothelial dysfunction: Common highway to type 2 diabetes and CVDCurr Diab Rep6279286200610.1007/s11892-006-0061-416879779 NitertMDDayehTVolkovPElgzyriTHallENilssonEYangBTLangSParikhHWessmanYImpact of an exercise intervention on DNA methylation in skeletal muscle from first-degree relatives of patients with type 2 diabetesDiabetes6133223332201210.2337/db11-1653230281383501844 ChristensenBCHousemanEAMarsitCJZhengSWrenschMRWiemelsJLNelsonHHKaragasMRPadburyJFBuenoRAging and environmental exposures alter tissue-specific DNA methylation dependent upon CpG island contextPLoS Genet5e1000602200910.1371/journal.pgen.1000602196804442718614 YoonGZhengYZhangZZhangHGaoTJoyceBZhangWGuanWBaccarelliAAJiangWUltra-high dimensional variable selection with application to normative aging study: DNA methylation and metabolic syndromeBmc Bioinformatics18156201710.1186/s12859-017-1568-1282646535340011 JungJJungYGillBKimCHwangKJJuYRLeeHJChuHHwangGSMetabolic responses to Orientia tsutsugamushi infection in a mouse modelPlos Negl Trop Dis9e3427201510.1371/journal.pntd.0003427255695624287389 HalstedCHMediciVAberrant hepatic methionine metabolism and gene methylation in the pathogenesis and treatment of alcoholic steatohepatitisInt J Hepatol2012959746201210.1155/2012/95974622007317 SvingenGFSchartum-HansenHPedersenERUelandPMTellGSMellgrenGNjølstadPRSeifertRStrandEKarlssonTNygårdOProspective associations of systemic and urinary choline metabolites with incident type 2 diabetesClin Chem62755765201610.1373/clinchem.2015.25076126980210 YangBTDayehTAKirkpatrickCLTaneeraJKumarRGroopLWollheimCBNitertMDLingCInsulin promoter DNA methylation correlates negatively with insulin gene expression and positively with HbA(1c) levels in human pancreatic isletsDiabetologia54360367201110.1007/s00125-010-1967-621104225 YangBTDayehTAVolkovPAKirkpatrickCLMalmgrenSJingXRenströmEWollheimCBNitertMDLingCIncreased DNA methylation and decreased expression of PDX-1 in pancreatic islets from patients with type 2 diabetesMol Endocrinol2612031212201210.1210/me.2012-1004225703315416998 ClineGWMagnussonIRothmanDLPetersenKFLaurentDShulmanGIMechanism of impaired insulin-stimulated muscle glucose metabolism in subjects with insulin-dependent diabetes mellitusJ Clin Invest9922192224199710.1172/JCI1193959151794508052 AsmannYWCarpenterJEShortKRStumpCSNairKSAltered skeletal muscle mitochondrial gene transcriptions in response to insulin infusion in T2D patientsMeeting of the American-Diabetes-Association532004 LoriaPLonardoAAnaniaFLiver and diabetes. A vicious circleHepatol Res435164201310.1111/j.1872-034X.2012.01031.x233320873733501 KirchnerHSinhaIGaoHRubyMASchönkeMLindvallJMBarrèsRKrookANäslundEDahlman-WrightKZierathJRAltered DNA methylation of glycolytic and lipogenic genes in liver from obese and type 2 diabetic patientsMol Metab5171183201610.1016/j.molmet.2015.12.004269773914770265 CarvalhoBSIrizarryRAA framework for oligonucleotide microarray preprocessingBioinformatics2623632367201010.1093/bioinformatics/btq431206889762944196 AssenovYMullerFLutsikPWalterJLengauerTBockCComprehensive analysis of DNA methylation data with RnBeadsNat Methods1111381140201410.1038/nmeth.3115252622074216143 DavisSDuPBilkeSTricheTBootwallaMMethylumi: Handle Illumina methylation dataR packageversion 2.24.12013 TricheTJJrWeisenbergerDJVan Den BergDLairdPWSiegmundKDLow-level processing of Illumina Infinium DNA Methylation BeadArraysNucleic Acids Res41e90201310.1093/nar/gkt090234760283627582 TeschendorffAEMarabitaFLechnerMBartlettTTegnerJGomez-CabreroDBeckSA beta-mixture quantile normalization method for correcting probe design bias in Illumina Infinium 450 k DNA methylation dataBioinformatics29189196201310.1093/bioinformatics/bts68023175756 ChenYALemireMChoufaniSButcherDTGrafodatskayaDZankeBWGallingerSHudsonTJWeksbergRDiscovery of cross-reactive probes and polymorphic CpGs in the Illumina Infinium HumanMethylation450 microarrayEpigenetics8203209201310.4161/epi.23470233146983592906 SmythGKlimma: Linear models for microarray data, In Bioinformatics and Computational Biology Solutions Using R and BioconductorGentlemanRCareyVDudoitSIrizarryRHuberWSpringerNew York, NY3974202005 KanehisaMGotoSKEGG: Kyoto encyclopedia of genes and genomesNucleic Acids Res282730200010.1093/nar/28.1.2710592173102409 HarrisMAClarkJIrelandALomaxJAshburnerMFoulgerREilbeckKLewisSMarshallBMungallCThe gene ontology (GO) database and informatics resourceNucleic Acids Res32(Database Issue)D258D261200410.1093/nar/gkh03614681407308770 Von MeringCHuynenMJaeggiDSchmidtSBorkPSnelBSTRING: A database of predicted functional associations between proteinsNucleic Acids Res31258261200310.1093/nar/gkg03412519996165481 ShannonPMarkielAOzierOBaligaNSWangJTRamageDAminNSchwikowskiBIdekerTCytoscape: A software environment for integrated models of biomolecular interaction networksGenome Res1324982504200310.1101/gr.123930314597658403769 HeXZhangJWhy do hubs tend to be essential in protein networks?PLoS Genet2e88200610.1371/journal.pgen.0020088167518491473040 JankyRSVerfaillieAImrichováHVan de SandeBStandaertLChristiaensVHulselmansGHertenKSanchezNaval MPotierDiRegulon: From a gene list to a gene regulatory network using large motif and track collectionsPLoS Comput Biol10e1003731201410.1371/journal.pcbi.1003731250581594109854 WangJDuncanDShiZZhangBWEB-based GEne SeT AnaLysis Toolkit (WebGestalt): Update 2013Nucleic Acids Res41W77W83201310.1093/nar/gkt439237032153692109 ZhangBKirovSSnoddyJWebGestalt: An integrated system for exploring gene sets in various biological contextsNucleic Acids Res33W741W748200510.1093/nar/gki475159805751160236 HobertOGene regulation by transcription factors and microRNAsScience31917851786200810.1126/science.115165118369135 ZhuoXZhangPHoergerTJLifetime direct medical costs of treating type 2 diabetes and diabetic complicationsAm J Prev Med45253261201310.1016/j.amepre.2013.04.01723953350 FuMWangCLiZSakamakiTPestellRGMinireview: Cyclin D1: Normal and abnormal functionsEndocrinology14554395447200410.1210/en.2004-095915331580 ChowdhuryMKMontgomeryMKMorrisMJCognardEShepherdPRSmithGCGlucagon phosphorylates serine 552 of β-catenin leading to increased expression of cyclin D1 and c-Myc in the isolated rat liverArch Physiol Biochem1218896201510.3109/13813455.2015.104869326135564 DaiCLiNSongGYangYNingXInsulin-like growth factor 1 regulates growth of endometrial carcinoma through PI3k signaling pathway in insulin-resistant type 2 diabetesAm J Transl Res8332933362016276481235009385 TaneeraJFadistaJAhlqvistEZhangMWierupNRenströmEGroopLExpression profiling of cell cycle genes in human pancreatic islets with and without type 2 diabetesMol Cell Endocrinol3753542201310.1016/j.mce.2013.05.00323707792 Karachanak-YankovaSDimovaRNikolovaDNeshevaDKoprinarovaMMaslyankovSTafradjiskaRGatevaPVelizarovaMHammoudehZEpigenetic alterations in patients with type 2 diabetes mellitusBalkan J Med Genet1815242016277853925026264 WalkerBAWardellCPChiecchioLSmithEMBoydKDNeriADaviesFERossFMMorganGJAberrant global methylation patterns affect the molecular pathogenesis and prognosis of multiple myelomaBlood117553562201110.1182/blood-2010-04-27953920944071 LiuCLinJDPGC-1 coactivators in the control of energy metabolismActa Biochim Biophys Sin (Shanghai)43248257201110.1093/abbs/gmr007213253363063079 SempleRKCrowleyVCSewterCPLaudesMChristodoulidesCConsidineRVVidal-PuigAO'rahillySExpression of the thermogenic nuclear hormone receptor coactivator PGC-1alpha is reduced in the adipose tissue of morbidly obese subjectsInt J Obes Relat Metab Disord28176179200410.1038/sj.ijo.080248214557831 RuschkeKFishbeinLDietrichAKlötingNTönjesAOberbachAFasshauerMJenknerJSchönMRStumvollMGene expression of PPARgamma and PGC-1alpha in human omental and subcutaneous adipose tissues is related to insulin resistance markers and mediates beneficial effects of physical trainingEur J Endocrinol162515523201010.1530/EJE-09-076719966034 HammarstedtAJanssonPAWesslauCYangXSmithUReduced expression of PGC-1 and insulin-signaling molecules in adipose tissue is associated with insulin resistanceBiochem Biophys Res Commun301578582200310.1016/S0006-291X(03)00014-712565902 PattiMEButteAJCrunkhornSCusiKBerriaRKashyapSMiyazakiYKohaneICostelloMSacconeRCoordinated reduction of genes of oxidative metabolism in humans with insulin resistance and diabetes: Potential role of PGC1 and NRF1Proc Natl Acad Sci USA10084668471200310.1073/pnas.103291310012832613166252 LingCDel GuerraSLupiRRönnTGranhallCLuthmanHMasielloPMarchettiPGroopLDel PratoSEpigenetic regulation of PPARGC1A in human type 2 diabetic islets and effect on insulin secretionDiabetologia51615622200810.1007/s00125-007-0916-5182706812270364 ChypreMZaidiNSmansKATP-citrate lyase: A mini-reviewBiochem Biophys Res Commun42214201210.1016/j.bbrc.2012.04.14422575446 GuayCMadirajuSRAumaisAJolyEPrentkiMA role for ATP-citrate lyase, malic enzyme and pyruvate/citrate cycling in glucose-induced insulin secretionJ Biol Chem2823565735665200710.1074/jbc.M70729420017928289 MenendezJVazquez-MartinAOrtegaFJFernandez-RealJMFatty acid synthase: Association with insulin resistance, type 2 diabetes, and cancerClin Chem55425438200910.1373/clinchem.2008.11535219181734 XiaYWanXDuanQHeSWangXInhibition of protein kinase B by palmitate in the insulin signaling of HepG2 cells and the preventive effect of archidonic acid on insulin resistanceFront Med China1200206200710.1007/s11684-007-0038-y24557677 CorderoPGome-ZurizAMCampionJMilagroFIMartinezJADietary supplementation with methyl donors reduces fatty liver and modifies the fatty acid synthase DNA methylation profile in rats fed an obesogenic dietGenes Nutr8105113201310.1007/s12263-012-0300-z22648174 DobrzynPJazurekMDobrzynAStearoyl-CoA desaturase and insulin signaling-what is the molecular switch?Biochim Biophys Acta179711891194201010.1016/j.bbabio.2010.02.00720153289 VossMDBehaATennagelsNTschankGHerlingAWQuintMGerlMMetz-WeidmannCHaunGKornMGene expression profiling in skeletal muscle of Zucker diabetic fatty rats: Implications for a role of stearoyl-CoA desaturase 1 in insulin resistanceDiabetologia4826222630200510.1007/s00125-005-0025-216284748 SchwenkRWJonasWErnstSBKammelAJähnertMSchürmannADiet-dependent alterations of hepatic Scd1 expression are accompanied by differences in promoter methylationHorm Metab Res45786794201310.1055/s-0033-134826323803969 KarahashiMHirata-HantaYKawabataKTsutsumiDKametaniMTakamatsuNSakamotoTYamazakiTAsanoSMitsumotoAAbnormalities in the metabolism of fatty acids and triacylglycerols in the liver of the goto-kakizaki rat: A model for non-obese type 2 diabetesLipids51955971201610.1007/s11745-016-4171-827372943 GermanMSGlucose sensing in pancreatic islet beta cells: The key role of glucokinase and the glycolytic intermediatesProc Natl Acad Sci USA9017811785199310.1073/pnas.90.5.1781844659145963 MatschinskyFMGlucokinase, glucose homeostasis, and diabetes mellitusCurr Diab Rep5171176200510.1007/s11892-005-0005-415929862 MullerYLPiaggiPHoffmanDHuangKGeneBKobesSThearleMSKnowlerWCHansonRLBaierLJBogardusCCommon genetic variation in the glucokinase gene (GCK) is associated with type 2 diabetes and rates of carbohydrate oxidation and energy expenditureDiabetologia5713821390201410.1007/s00125-014-3234-8247281274052004 TangLYeHHongQWangLWangQWangHXuLBuSZhangLChengJElevated CpG island methylation of GCK gene predicts the risk of type 2 diabetes in Chinese malesGene547329333201410.1016/j.gene.2014.06.06224992032 ValenteABichoMDuarteRRaposoJFCostaHSDietary sodium intake related with cysteine and methionine in type 2 diabetic patientsAtherosclerosis235e108e109201410.1016/j.atherosclerosis.2014.05.293 ChenHJYangYFLaiPYChenPFAnalysis of chlorination, nitration, and nitrosylation of tyrosine and oxidation of methionine and cysteine in hemoglobin from type 2 diabetes mellitus patients by nanoflow liquid chromatography tandem mass spectrometryAnal Chem8892769284201610.1021/acs.analchem.6b0266327541571 VelmuruganGVWhiteCCystathionine gamma-lyase deficiency impairs H2S biosynthesis and vessel reactivity in type-2 diabetesFASEB J271 SupplS1091S10932013 HottaNKakutaHFukasawaHKohNSakakibaraFNakamuraJHamadaYWakaoTHaraTMoriKEffect of a potent new aldose reductase inhibitor, (5-(3-thienyltetrazol-1-yl)acetic acid (TAT), on diabetic neuropathy in ratsDiabetes Res Clin Pract27107117199510.1016/0168-8227(95)01033-A7607048

Functional and pathway enrichment analyses of upregulated and downregulated genes in patients with T2D. The results of pathway analysis for (A) downregulated genes and (B) upregulated genes in patients with T2D. The top five enriched GO terms for (C) downregulated genes and (D) upregulated genes in patients with T2D. KEGG, Kyoto Encyclopedia of Genes and Genomes; GO, Gene Ontology; BP, biological process; MF, molecular function; CC, cellular component; NADP (H), nicotinamide adenine dinucleotide phosphate; T2D, type 2 diabetes.

Protein-protein interaction network for the differentially-expressed genes. Red circle nodes represent upregulated genes and the green diamond nodes represent downregulated genes. Overlapping genes between the differentially-expressed and -methylated genes are indicated by surrounding blue borders.

Transcriptional factor target network. White hexagonal nodes represent the predicted transcription factors. The red circle nodes represent upregulated genes and the green diamond nodes represent downregulated genes. Overlapping genes between the differentially-expressed and -methylated genes are indicated by surrounding blue borders.

MicroRNA target network. White triangle nodes represent the predicted microRNAs. The red cycle nodes represent upregulated genes and the green diamond nodes represent downregulated genes. Overlapping genes between the differentially-expressed and -methylated genes are indicated by surrounding blue borders.

KEGG pathways and the top 10 GO BP terms enriched by the overlapping genes.

A, Pathway
ID Name Count P-value Genes
hsa00330 Arginine and proline metabolism 4 5.53×10−3 GLS2, ALDH18A1, OAT, PRODH
B, Biological process
ID Name Count P-value Genes
GO:0016053 Organic acid biosynthetic process 7 3.07×10−4 ALDH18A1, SDS, SCD, ELOVL2, FASN, LGSN, PRODH
GO:0046394 Carboxylic acid biosynthetic process 7 3.07×10−4 ALDH18A1, SDS, SCD, ELOVL2, FASN, PRODH
GO:0055114 Oxidation reduction 13 3.52×10−4 ME1, HSD17B11, TP53I3, ALDH18A1, FMO1, SCD, CYP4F22, FASN, AASS, CYP26A1, PPARGC1A, HPGD, PRODH
GO:0009064 Glutamine family amino acid metabolic process 4 3.57×10−3 GLS2, ALDH18A1, LGSN, PRODH
GO:0006739 NADP metabolic process 3 3.93×10−3 ME1, TP53I3, GCK
GO:0009084 Glutamine family amino acid biosynthetic process 3 5.53×10−3 ALDH18A1, LGSN, PRODH
GO:0033273 Response to vitamin 4 6.97×10−3 CCND1, PDGFA, IGFBP2, SPP1
GO:0007156 Homophilic cell adhesion 5 7.53×10−3 RET, CDH15, FAT1, DSG1, CDH23
GO:0007155 Cell adhesion 11 8.04×10−3 RET, CDH15, EPDR1, LAMA5, FAT1, DSG1, CPXM2, IL32, CYR61, SPP1, CDH23
GO:0022610 Biological adhsion 11 8.12×10−3 RET, CDH15, EPDR1, LAMA5, FAT1, DSG1, CPXM2, IL32, CYR61, SPP1, CDH23

KEGG, Kyoto Encyclopedia of Genes and Genomes; GO, Gene Ontology; BP, biological process.

Top 20 genes with higher degrees in the protein-protein interaction network.

Genes Degree
ACLY 19
FASN 15
CCND1 13
GCK 12
SCD 12
MET 12
IGF1 12
ALDH18A1 11
SERPINE1 9
FDFT1 8
SQLE 8
TAT 7
HSPA5 7
IGFBP1 7
CDKN1A 7
C10orf10 6
SDS 6
SPP1 6
FABP4 6
PPARGC1A 6

KEGG pathways and GO BP terms enriched for the top 20 nodes in the protein-protein interaction network.

A, Pathway
Pathway ID Pathway name Count P-value Genes
hsa04115 p53 signaling pathway 4 1.36×10−4 CDKN1A, CCND1, SERPINE1, IGF1
hsa05218 Melanoma 4 1.54×10−3 CDKN1A, CCND1, MET, IGF1
hsa05214 Glioma 3 1.82×10−2 CDKN1A, CCND1, IGF1
hsa04510 Focal adhesion 4 2.75×10−2 CCND1, MET, IGF1, SPP1
hsa05215 Prostate cancer 3 3.47×10−2 CDKN1A, CCND1, IGF1
B, Biological process
Pathway ID Pathway name Count P-value Genes
GO:0031099 Regeneration 5 1.80×10−6 CDKN1A, CCND1, SERPINE1, IGF1, IGFBP1
GO:0010033 Response to organic substance 8 2.25×10−5 CDKN1A, CCND1, GCK, SQLE, FABP4, IGFBP1, TAT, SPP1
GO:0009725 Response to hormone stimulus 6 9.15×10−5 CDKN1A, CCND1, FABP4, IGFBP1, TAT, SPP1
GO:0048545 Response to steroid hormone stimulus 5 1.03×10−4 CDKN1A, CCND1, FABP4, TAT, SPP1
GO:0051384 Response to glucocorticoid stimulus 4 1.41×10−4 CDKN1A, CCND1, FABP4, TAT
GO:0009719 Response to endogenous stimulus 6 1.46×10−4 CDKN1A, CCND1, FABP4, IGFBP1, TAT, SPP1
GO:0009991 Response to extracellular stimulus 5 1.74×10−4 CDKN1A, CCND1, HSPA5, PPARGC1A, SPP1
GO:0031960 Response to corticosteroid stimulus 4 1.82×10−4 CDKN1A, CCND1, FABP4, TAT
GO:0010907 Positive regulation of glucose metabolic process 3 2.82×10−4 GCK, IGF1, PPARGC1A
GO:0010676 Positive regulation of cellular carbohydrate metabolic process 3 3.13×10−4 GCK, IGF1, PPARGC1A

KEGG, Kyoto Encyclopedia of Genes and Genomes; GO, Gene Ontology.

Top 12 genes with higher degrees in the transcriptional factor-target regulatory network.

Genes Degree
CCND1 15
PPARGC1A 11
WNT11 11
ZSCAN16 10
PRKCE 9
SULF1 9
GCK 8
RET 8
SLITRK3 8
OAT 7
SCD 7
ME1 7

Top 14 genes with higher degrees in the microRNA target regulatory network.

Genes Degree
TP53INP1 15
SLC1A2 13
PPARGC1A 12
SLITRK3 11
L3MBTL3 11
RET 10
CCND1 8
RBMS1 7
GRID1 7
PSD3 7
PRKCE 6
ANKS1B 5
SULF1 5
HMCN1 5

Overlapped genes between the GSE64998 dataset and GSE15653 validation dataset.

Common upregulated genes Common downregulated genes
RDH16, SLC39A7, KRT8, AEN, HSPA5, KDM8, ALDH18A1, ATF5, CDHR2, CCND1, SLC35C1, CD151, GAS6, ACOT1, HPS5, APOL3, FASN, CHI3L1, CDKN1A, HBB SLITRK3, CYP2C19, SLC16A4, CA14, GPR88, OPN3, VIL1, HAL, HERC5, DNAJC12, TAT, PFKFB3, SMPDL3A, SLC19A2, ABCA8