Table 1.
Characteristics of EWASs associated with type 2 diabetes
| Referencea | Population for DNA methylation analysis | Female/male | Design | Tissue | Method | Covariates included in analysis | Multiple-testing correctionb | Top findings | No. of CpGs included in replication studyc |
|---|---|---|---|---|---|---|---|---|---|
| Chambers et al, 2015 [21]a | 1074 incident type 2 diabetes patients, 1590 controls (Indian Asians); 1141 Europeans for replication |
P 352/722 C 426/1083 |
Longitudinal (nested case–control) | Blood | 450k for discovery; pyrosequencing and 450k for replication | Age, sex, intensity values from 450k control probes, batch, measured and imputed (Houseman method) cell count, 5 PC | Discovery p < 5 × 10−7
Meta-analysis Bonferroni |
5 DMS replicated in independent cohorts: TXNIP gene (cg19693031); SREBF1 gene (cg11024682); PHOSPHO1 gene (cg02650017); SOCS3 gene (cg18181703); ABCG1 gene (cg06500161) | 5 |
| Al Muftah et al, 2016 [22]a | 30 T2D patients, 93 controls from 15 families of Qatari descent; 810 female twins from TwinsUK for replication |
72/51 | Cross-sectional (case–control) | Blood | 450k | Age, sex, smoking status, cell count (Houseman method); BMI as confounder | Bonferroni | 1 DMS: DQX1 gene (cg06721411) replicated in TwinsUK | 1 |
| Soriano-Tarraga et al, 2016 [23]a | 151 T2D patients and 204 controls from IS cohort; BISMAR_2 (59 T2D patients and 108 controls); REGICOR (63 T2D patients and 582 controls) for replication |
P 61/90 C 88/116 |
Cross-sectional (case–control) | Blood | 450k | Age, sex, smoking, hyperlipidaemia, BMI, cell count (Houseman method) | Bonferroni | 1 DMS: TXNIP gene (cg19693031) replicated in 2 independent cohorts | 1 |
| Florath et al, 2016 [24]a | 153 T2D patients and 835 controls; 87 T2D patients and 527 controls for replication |
P 56 & 59% male (controlled or poorly controlled T2D) C 49% male |
Cross-sectional (case–control) | Blood | 450k | Age, sex, BMI, smoking, batch, cell count (Houseman method) | FDR < 5% Bonferroni for replication |
39 DMS associated with T2D in discovery cohort, after replication in independent cohort 1 DMS remain significant: TXNIP (cg19693031) | 1 |
| Kulkarni et al, 2015 [25]a | 850 pedigreed Mexican Americans (174 T2D patients) | 536/314 | Cross-sectional (case–control) | Blood | 450k | Age, sex, BMI, cell count (Jaffe method) | Bonferroni | Overall 51 DMS associated with T2D status; 19 with fasting glucose level and 24 with HOMA-IR | 51 |
| Yuan et al, 2014 [26]a | 27 monozygotic twin pairs from TwinsUK (17 pairs T2D-discordant, 3 pairs T2D concordant and 7 healthy pairs); 42 unrelated T2D cases and 221 controls for replication | 23 pairs/4 pairs | Cross-sectional (twins study) | Blood (white blood cells) | MeDIP-seq 450k for replication |
Age, sex, BMI, beadchip, bisulphite conversion efficiency, (family as a random effect) | FDR < 5% | 1 DMR: MALT1 gene (ch18:56336501-56337000), replicated using Illumina 450k array (cg24182998), replicated reached Bonferroni threshold (0.05/20 = 0.0025) | 1 |
| Matsha et al, 2016 [13] | 3 T2D patients, 3 pre-diabetes, 3 controls (age, sex, BMI and duration of residence matched) | All female | Cross-sectional (case–control) | Blood | MeDIP sequencing | – | q = 10−2 | 450,142 DMRs were statistically significant in all samples, among others associated with cell surface receptor signalling, inflammatory pathways, glucose transport, muscle and pancreas development genes, insulin signalling | 0, not an Illumina array |
| Toperoff et al, 2012 [15] | Ashkenazi Jews: 710 T2D patients and 459 controls were assembled in 4 age-matched pools | Males fraction around 50% in all 4 pools | Cross-sectional with multistep design (case–control) | Blood | Microarray-based assay for methylation levels; sequencing of bisulfite converted DNA pools | Sex and lymphocyte percentage | FDR | 6 DMRs were found in LD blocks. After replication and multiple hypothesis testing 13 out of 93 CpGs located in 6 DMRs showed significant case–control difference. Among them the nearest genes were: THADA, JAZF1, SLC30A8, TCF7L2, KCNQ1 and FTO. CpG site near FTO showed small (3.35%) but significant hypomethylation of cases vs controls | 0, not an Illumina array |
| Bacos et al, 2016 [12] | 87 non-diabetic donors for discovery phase and 112 individuals from Danish Family Study | 34/53 | Cross-sectional association with age | Pancreatic islets and whole blood | 450k genome-wide and 4 sites replicated using pyrosequencing in blood | Sex, BMI, HbA1c, bisulphite treatment, days in culture and islet purity | FDR < 5% | Ageing was significantly associated with altered DNA methylation at 241 sites in pancreas; almost 60% of sites were found also in other studies in blood; 4 most significant sites (FHL2, ZNF518B, GNPNAT1 and HLTF) were selected for follow-up analysis and they showed functional effects on beta cells or association with T2D risk. Higher methylation of those sites was associated with lower risk of T2D development during progression into T2D (mean 10.8 years; Botnia Prospective Study) | 0, indirect association with diabetes |
| Volkov et al, 2017 [14] | 6 T2D donors and 8 control donors | P 3/3 C 4/4 |
Cross-sectional (case–control) | Pancreatic islets | WGBS | – | – | Average methylation level was 75.9%, 25,820 DMRs were identified in T2D pancreatic islets, while 13,696 were hypermethylated and 12,124 were hypomethylated. 692 DMRs had a methylation difference > 10%, the highest in regions annotated to ARX and TFAM genes | 0, not an Illumina array |
| Dayeh et al, 2014 [27]a | 15 T2D donors and 34 control donors | P 5/10 C 12/22 |
Cross-sectional (case–control) | Pancreatic islets | 450k | Age, sex, BMI, batch, islet purity, day of culture | FDR < 5% | 1649 DMS (853 genes and 561 intragenic) with at least 5% difference in methylation between diabetic and non-diabetic donors | 15 |
| Volkmar et al, 2012 [28]a | 5 T2D donors and 11 non-diabetic donors matched by age, BMI and ethnicity; 12 T2D patients and 12 age- and BMI-matched controls for replication |
– | Cross-sectional (case–control) | Pancreatic islets | 27k | – | p < 0.01 and 15% group-wise difference on methylation level | 276 DMS (254 genes) were found, 96% were hypomethylated | 0 |
| Nilsson et al, 2014 [29]a | 14 monozygotic twins discordant for T2D; Cohort 2: 28 T2D patients/28 controls (unrelated) | P 5/9 C 5/9 |
Cross-sectional (twins study) | Adipose tissue | 450k | BMI, glucose | FDR < 15% | In twins 23,470 DMS were found, none passed FDR correction In Cohort 2 15,627 DMS (7046 genes) were found after a FDR correction DNA methylation of 266 sites, corresponding to 103 genes, was significantly associated with expression in the discordant twins at q < 0.15 |
0 |
| Ribel-Madsen et al, 2012 [30]a | 12 Danish monozygotic twins discordant for T2D | P 6/6 C 6/6 |
Cross-sectional (twins study) | Skeletal muscle (11 pairs) Adipose tissue (5 pairs) |
27k | – | Padj (Westfall–Young resampling method p < 0.001) | 1 DMS in skeletal muscle: IL8 gene and 7 DMS in adipose tissue: ZNF668; HSPA2; C8orf31; CD320; TWIST1; MYO5A | 0 |
| Kirchner et al, 2016 [31]a | 8 obese T2D men, 7 obese non-diabetic controls and 7 non-obese metabolically healthy control individuals | All male | Cross-sectional (case–control) | Liver | 450k | – | FDR < 25% | 2255 DMS (1388 genes) were found (T2D obese compared with non-obese control individuals) | 0 |
| Nilsson et al, 2015 [32]a | 35 T2D patients and 60 control individuals | P 18/17 C 43/17 |
Cross-sectional (case–control) | Liver | 450k | Age, sex, BMI, NASH diagnosis, degree of steatosis | FDR < 5% | 251 DMS (167 genes) were found mostly hypomethylated (94%) in those with T2D. A decrease in folate levels in T2D patients was observed, which could explain decreased methylation in the human liver in diabetes | 3 |
aStudies included into replication study to provide CpGs, which passed strict Bonferroni correction threshold
bMultiple-testing threshold calculated originally by the authors of particular study
c0 in last column means zero CpG sites passed study-specific Bonferroni correction threshold
C, controls; DMS, differentially methylated sites; DMR, differentially methylated region; FDR, false discovery rate; HOMA-IR, homeostatic model assessment; IS, ischemic stroke; LD, linkage disequilibrium; NASH, non-alcoholic steatohepatitis; P, patients; PC, principal component; T2D, type 2 diabetes; 27k, Infinium HumanMethylation27 BeadChip; 450k, Infinium HumanMethylation450 BeadChip