Skip to main content
NCBI home page
Oxford University Press logoLink to Oxford University Press
. 2020 Feb 4;65(11):1414–1425. doi: 10.1373/clinchem.2019.305508

5-Hydroxymethylcytosines in Circulating Cell-Free DNA Reveal Vascular Complications of Type 2 Diabetes

Ying Yang 1,†, Chang Zeng 2,3,†,✉, Xingyu Lu 4, Yanqun Song 4, Ji Nie 5, Ruoxi Ran 1, Zhou Zhang 3,✉, Chuan He 5,6,✉,‡, Wei Zhang 3,✉,‡, Song-Mei Liu 1,✉,‡
PMCID: PMC7055673  PMID: 31575611

Abstract

BACKGROUND

Long-term complications of type 2 diabetes (T2D), such as macrovascular and microvascular events, are the major causes for T2D-related disability and mortality. A clinically convenient, noninvasive approach for monitoring the development of these complications would improve the overall life quality of patients with T2D and help reduce healthcare burden through preventive interventions.

METHODS

A selective chemical labeling strategy for 5-hydroxymethylcytosines (5hmC-Seal) was used to profile genome-wide 5hmCs, an emerging class of epigenetic markers implicated in complex diseases including diabetes, in circulating cell-free DNA (cfDNA) from a collection of Chinese patients (n = 62). Differentially modified 5hmC markers between patients with T2D with and without macrovascular/microvascular complications were analyzed under a case–control design.

RESULTS

Statistically significant changes in 5hmC markers were associated with T2D-related macrovascular/microvascular complications, involving genes and pathways relevant to vascular biology and diabetes, including insulin resistance and inflammation. A 16-gene 5hmC marker panel accurately distinguished patients with vascular complications from those without [testing set: area under the curve (AUC) = 0.85; 95% CI, 0.73–0.96], outperforming conventional clinical variables such as urinary albumin. In addition, a separate 13-gene 5hmC marker panel could distinguish patients with single complications from those with multiple complications (testing set: AUC = 0.84; 95% CI, 0.68–0.99), showing superiority over conventional clinical variables.

CONCLUSIONS

The 5hmC markers in cfDNA reflected the epigenetic changes in patients with T2D who developed macrovascular/microvascular complications. The 5hmC-Seal assay has the potential to be a clinically convenient, noninvasive approach that can be applied in the clinic to monitor the presence and severity of diabetic vascular complications.

Type 2 diabetes (T2D)7 is a formidable pandemic challenge to human health. More than 424 million people are living with diabetes globally, and the number is expected to increase to 628 million by 2045 (1). China has the largest diabetic population, estimated at approximately 114.4 million in 2017 (1). The heaviest global burden of T2D arises from complications that increase with duration of disease and result in disability and reduced life expectancy (2). Patients with T2D are at particularly high risk of vascular complications (3, 4), which are traditionally divided into macrovascular (i.e., atherosclerotic diseases, heart diseases, and stroke) and microvascular complications (i.e., nephropathy, neuropathy, retinopathy) (3, 5). An observational study of 66726 patients with T2D from Asia, Africa, Europe, and South America revealed that vascular complications were common, with 27.2% and 53.5% of patients suffering from macrovascular and microvascular complications, respectively (6). Notably, in patients with T2D, atherosclerosis is the main cause of impaired life expectancy, and diabetic nephropathy and retinopathy are the predominant contributors to the end-stage renal disease and blindness, respectively (5). Currently, no consensus exists in the classification, definition, or diagnostic criteria for the complications of T2D (5). Novel approaches including effective biomarkers are expected to improve the diagnosis of T2D complications as well as predict the clinical outcomes.

In mammals, 5-methylcytosines (5mCs) in DNA are dynamic and reversible (7, 8) and can be oxidized into 5-hydroxymethylcytosines (5hmCs) and other modified cytosine intermediates through the Ten–Eleven Translocation enzymes (9). Notably, 5hmCs have been shown to be stable epigenetic marker in the human genome, not simply the intermediate products of 5mC demethylation. The 5hmCs vary from tissue to tissue, showing enrichment in brain, liver, kidney, and colorectal tissue while being at very low levels in heart, breast, and placenta (10). In contrast to the suppressive effect of 5mCs at promoters, 5hmCs are widely regarded as markers of activated genic loci or chromatin (11). 5hmCs play vital roles in cell development, differentiation, maturation, and self-renewal (12). Reduction of global levels of 5hmCs has been reported in various cancers and nervous system disorders, indicating their relevance to complex diseases (12–15). Recently, studies have suggested correlation between alterations in 5hmCs and diabetic complication. For instance, a recent study using retinal capillary cells and retina from diabetic mice indicated that hyperglycemia induced matrix metalloproteinase-9 promoter hypomethylation and transcription enhancement, and silencing of Ten–Eleven Translocation 2 could prevent this phenomenon (16). Moreover, peripheral blood levels of 5hmCs were increased in patients with T2D who had glucose levels that were poorly controlled compared to patients with well-controlled glucose levels and healthy individuals (17). Periods of persistent hyperglycemia during the course of T2D predispose the patient to increased risk of developing diabetic complications (17). These characteristics revealed potential diagnostic or predictive value of 5hmCs for diabetic complications.

Cell-free DNA (cfDNA) in the circulating blood originates from dying cells from different tissues, which release DNA into the peripheral bloodstream upon degradation after cell death (18). CfDNA in blood has attracted great interest as a noninvasive source for the clinic with potential roles in disease diagnosis, such as cancers and cardiovascular diseases (15, 19, 20). However, conventional epigenomic profiling approaches, e.g., bisulfite conversion–based methods, are limited with cfDNA degradation, and they cannot distinguish 5hmC from 5mC. Here we developed a selective chemical labeling strategy for 5hmCs, the 5hmC-Seal (21), which was optimized for genome-wide 5hmC mapping in as low as approximately 1–2 ng cfDNA from only a few milliliters of plasma (19). Taking advantage of this technique, we profiled 5hmCs in cfDNA samples derived from a collection of patients with T2D with or without vascular complications. Using a case–control design, we evaluated whether 5hmC markers in cfDNA could distinguish patients who had developed macrovascular (and/or microvascular) complications and who did not yet manifest obvious signs of these complications.

Materials and Methods

STUDY PARTICIPANTS

In total, 62 patients with T2D were recruited from the Specialist Clinic of the Department of Endocrinology at Zhongnan Hospital, Wuhan University, China. Fasting blood samples and 24-h urine collections were obtained for the clinical tests and plasma cfDNA extraction on the morning following hospital admission. The clinical information of patients was collected from the medical records following a standard protocol (see Table 1 in the Data Supplement that accompanies the online version of this article at http://www.clinchem.org/content/vol65/issue11). Patients with T2D were diagnosed according to the 2017 Standards of Medical Care in Diabetes of the American Diabetes Association (22). All patients underwent vascular ultrasound, formal eye examinations, and routine clinical laboratory tests. Among these patients, 12 had T2D without vascular complications and 50 were diagnosed as having microvascular and/or macrovascular complications (Fig. 1A; see Table 1 in the online Data Supplement). Patients with T2D who had atherosclerotic plaque or a clinical history of atherosclerotic heart disease or stroke were defined as having macrovascular complications, and patients with nephropathy, neuropathy, or retinopathy were defined as having microvascular complications (3, 5). The presence or absence of atherosclerotic plaque in the carotid artery and/or lower limb artery was assessed by the Logiq E9 General Imaging-ultrasound System (General Electric). Nephropathy was defined as the presence of albuminuria (>20 μg/min) and/or reduced estimated glomerular filtration rate (eGFR; <60 mL/min/1.73m2) (23). Neuropathy was assessed by clinical symptoms such as unpleasant sensations of burning, tingling, and skin itching, as well as the quantitative sensory testing (23). Retinopathy was evaluated by a study ophthalmologist using the YZ6H ophthalmoscope (66 Vision) with pupil dilation and the CIRRUS HD-OCT 5000 optical coherence tomography (Zeiss) according to typical retinal microvascular signs (see Methods file in the online Data Supplement) (23, 24). Patients showing only 1 type of these complications were defined as having single complications and the patients showing 2 or more types were defined as having multiple complications. All samples were deidentified before cfDNA extraction, 5hmC profiling, and statistical modeling. The Ethics Committee of Zhongnan Hospital approved this study. Informed consent was obtained from each study participant.

Fig. 1. Overview of the 5hmC-Seal and study design.

Fig. 1.

Open in a new tab

(A), The concept of distinguishing patients with T2D who had complications from patients without complications using the 5hmC-Seal technique, an approach utilizing clinically convenient liquid biopsy. (B), Flowchart showing the general design of the study, which aimed to detect 5hmC-based marker genes for patients with T2D who had vascular complications in plasma cfDNA. NGS, next-generation sequencing.

CLINICAL LABORATORY TESTS

Liver function tests [serum alanine transaminase (ALT), aspartate transaminase (AST), AST/ALT, total bilirubin, direct bilirubin, indirect bilirubin, total protein, albumin, globin, γ-glutamyl transpeptidase, alkaline phosphatase, and total bile acid], renal function tests (serum urea nitrogen, creatinine, uric acid, eGFR), serum glucose, and lipids (total cholesterol, triglycerides, HDL cholesterol, LDL cholesterol, apolipoprotein A1, apolipoprotein B, lipoprotein A, nonesterified fatty acid) were determined with the AU5800 Chemistry Analyzer (Beckman). Serum insulin was assayed by the i4000SR Immunology Analyzer (Abbott Laboratories). Blood glycated hemoglobin (Hb A1c) was measured by the HA-8160 Glycohemoglobin Analyzer (ADAMS). The 24-h urine collections were used for screening albuminuria with the Immulite 1000 Immunoassay System (Siemens).

The eGFR was calculated by the Chronic Kidney Disease Epidemiology Collaboration equation: eGFR = 141 × min (Scr/κ, 1)α × max (Scr/κ, 1)–1.209 × 0.993Age × (1.018 if female), (Scr, Serum creatinine; unit, mg/dL, 1 mg/dL = 88.4 μmol/L), where κ = 0.7 for females or 0.9 for males; α = −0.329 for females or −0.411 for males; min indicates the minimum of Scr/κ or 1; and max indicates the maximum of Scr/κ or 1 (25).

5hmC-SEAL PROFILING

Plasma cfDNA samples were prepared from approximately 4 mL of peripheral blood with EDTA anticoagulation, as previously described (19). Details about the 5hmC-Seal library preparation and bioinformatic processing pipeline are described in previous publications (19, 21, 26) as well as in the Methods file in the online Data Supplement. The raw and processed 5hmC-Seal data have been deposited into the NCBI Gene Expression Omnibus database (GSE125929).

IDENTIFYING 5hmC MARKERS FOR VASCULAR COMPLICATIONS

We used a 2-step approach to develop a weighted model (score) for detecting vascular complications in patients with T2D (Fig. 1B). In Step 1, the normalized 5hmC-Seal data were used to identify a list of most informative candidate marker genes. In Step 2, the candidate marker genes from Step 1 were further selected to build a final panel of 5hmC marker genes by applying the elastic net regularization in a multivariable logistic regression model (see Methods file in the online Data Supplement).

EVALUATION OF THE 5hmC MARKERS AND CONVENTIONAL CLINICAL VARIABLES

Clinical variables were categorized into “normal” and “abnormal” according to the corresponding clinical reference intervals. Each clinical variable was used to assess the detection accuracy of vascular complications for all samples (27). In addition, we conducted multivariable analysis to analyze combinations of conventional clinical variables and the weighted scores (see Methods file in the online Data Supplement).

EXPLORING FUNCTIONAL RELEVANCE OF THE MARKER GENES

We used the Reactome Knowledgebase (28) to explore whether certain pathways were enriched in our candidate genes showing differential modification of 5hmC from Step 1. We used the Reactome Functional Interaction plug-in for Cytoscape to explore functional interactions across the detected marker genes and/or differentially modified candidates from Step 1. Hubs of the Reactome Functional Interaction networks were estimated based on the measurement of betweenness centrality, which represented the magnitude of influence a node (i.e., gene) had over the flow of information in a gene network. To investigate 5hmC modification levels and diabetic duration, we used the Cox proportional hazards model in the R package survivalAnalysis (29). Multivariate Cox proportional hazards model were built with T2D complications as dependent variables and age, sex, body mass index (BMI), and 5hmC levels of the candidate genes as independent variables. Here, age refers to age of the patients at baseline. The time-dependent hazard ratios (HRs) and 95% CIs were estimated.

Results

CLINICAL CHARACTERISTICS OF THE STUDY PARTICIPANTS

Table 1 in the online Data Supplement shows the clinical characteristics of the 62 study participants, including 12 patients without vascular complications, 34 patients with single vascular complications (24 cases with macrovascular events, 10 cases with microvascular events), and 16 patients with multiple vascular complications. In total, 36 patients used insulin treatment, 34 patients used oral glucose-lowering medications, 25 patients used blood pressure-lowering medications, and 7 patients used lipid-lowering agents. No significant difference of the concomitant medications was found among different groups of patients (P > 0.05; see Table 1 in the online Data Supplement). The mean age for the 62 patients was 59.1 ± 11.3 years (range, 34–86 years) and 61.3% (n = 38) were men. For the patients with complications, the mean diabetes duration was 8.6 ± 7.7 years (range, 0.02–30 years) at the time of sample collection, and 56% (n = 28) were men. Baseline characteristics of age, sex distribution, BMI, and smoking did not significantly differ between patients with complications and patients without complications or between patients with single complications and patients with multiple complications (P > 0.05; see Table 1 in the online Data Supplement).

5hmC MODIFICATION LEVELS ASSOCIATED WITH VASCULAR COMPLICATIONS

The general features of the profiled 5hmC markers in cfDNA are shown in Fig. 2. Overall 5hmC levels over gene body did not differ between male and female individuals. The 5hmC modification levels among the 62 patients with or without complications were compared using a logistic regression model controlled for sex and age. The normalized mean counts of the 662 differentially modified candidate genes (including both up- and down-modified genes) significantly increased in patients with complications relative to patients without complications (P < 0.001), but the mean counts had no statistically significant difference between patients with single complication and patients with multiple complications (Fig. 2A and B). The 5hmC modification landscapes among the 50 patients with single or multiple complications were then compared using a logistic regression model. The normalized mean counts of the 153 differentially modified candidate genes (including both up- and down-modified genes) decreased in patients with multiple complications (P < 0.001; Fig. 2D).

Fig. 2. Genomic landscapes of the 5hmC modifications in cfDNA.

Fig. 2.

Open in a new tab

(A), Patients with or without complications show distinct 5hmC modification landscapes based on the normalized counts of 662 candidate marker genes. (B), The mean of normalized counts of the 662 candidate marker genes among the analyzed patients with T2D. (C), Patients with single or multiple complications show distinct 5hmC modification landscapes based on the normalized counts of 153 candidate marker genes. (D), The mean of normalized counts of the 153 candidate marker genes differ significantly between patients with single or multiple complications. Wilcoxon Rank Sum Test P value: *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Abbreviations: T2D Comp (−), patients without complications; T2D Comp (+), patients with complications.

5hmC MARKERS FOR PATIENTS WITH OR WITHOUT COMPLICATIONS

The 62 patients were then randomly grouped into a training set (n = 31) and a testing set (n = 31). Our primary statistical modeling was carried out on approximately 20000 gene bodies with a priori determined criteria (i.e., at least 20 reads in >90% samples). Logistic regression models were used to select a list of 135 most informative candidate genes for further variable selection and weighted score computation (see Table 2 in the online Data Supplement). A 16-gene panel of 5hmC markers was identified using the elastic net for distinguishing patients with T2D with complications from those without (Fig. 3A). The weighted score for detecting complications was then computed based on the 16-gene panel and the final multivariable logistic model (Detection Model 1) from the training set. The patients without complications showed significantly lower scores than those with complications in both the training set and the testing set (Fig. 3B). The receiver operating characteristic curves were used to evaluate model performance by the area under the curve (AUC). Considering weighted score alone, the results indicated 60% clinical sensitivity and 100% clinical specificity at the score cutoff of 0.50 for patient classification in the testing set.

Fig. 3. Performance of the 5hmC markers in cfDNA for patients with T2D who had complications.

Fig. 3.

Open in a new tab

(A), The heatmaps and (B), the weighted scores of the training and testing groups showed the performance of the 16 marker genes for distinguishing patients withT2D who had complications from patients without complications. (C), The heatmaps and (D), the weighted scores of the training and testing groups show the performance of the 13 marker genes for distinguishing T2D patients with single complications from those with multiple complications. Wilcoxon rank sum test P value: *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Abbreviations: Nephropathy (–), patients with T2D without nephropathy; Nephropathy (+), patients with T2D with nephropathy. Human genes: CCDC74B, coiled-coil domain-containing protein 74B; FAM136A, family with sequence similarity 136 member A; GGACT, gamma-glutamylamine cyclotransferase; CCER2, coiled-coil glutamate rich protein 2; FAM19A3, TAFA chemokine like family member 3; SLC22A7, solute carrier family 22 member 7; COL1A2, collagen type I alpha 2 chain; RPL22L1, ribosomal protein L22 like 1; FGFBP1, fibroblast growth factor binding protein 1; ZNF391, zinc finger protein 391; ALG10B, ALG10 alpha-1,2-glucosyltransferase B; FAM162B, family with sequence similarity 162 member B; DEFB125, defensin beta 125; OR2L13, olfactory receptor family 2 subfamily L member 13; SLC22A24, solute carrier family 22 member 24; C6orf58, chromosome 6 open reading frame 58; PCDH18, protocadherin 18; IL17F, interleukin 17F; SPINK14, serine peptidase inhibitor, Kazal type 14 (putative); DEFB108B, defensin beta 108B; ZNF705E, zinc finger protein 705E.

5hmC MARKERS FOR PATIENTS WITH SINGLE OR MULTIPLE COMPLICATIONS

The 50 patients with available diagnostic information (i.e., with single or multiple complications) were randomly grouped into a training set (n = 25) and a testing set (n = 25). In total, 159 candidate marker genes were selected for further feature selection (see Table 3 in the online Data Supplement). A 13-gene panel of 5hmC marker genes was identified using the elastic net regularization for distinguishing patients with T2D with a single complication from those with multiple complications (Fig. 3C). The weighted score for detecting single/multiple complications was computed based on this 13-gene model (Detection Model 2). The weighted scores of patients with single complications were significantly different from the scores of those with multiple complications in both the training and testing sets (Fig. 3D). Notably, the weighted scores of patients with nephropathy were also different from the scores of those without nephropathy (Fig. 3D). When using the weighted score as the only predictor, the receiver operating characteristic curve results showed an 87.5% clinical sensitivity and 68.8% clinical specificity at the score cutoff of 0.69 for patient classification in the testing set.

COMPARISONS WITH CONVENTIONAL CLINICAL VARIABLES OR RISK FACTORS

We evaluated the sensitivity and specificity of various clinical variables or risk factors for detecting vascular complications in the study participants (Table 1). In general, the results suggested that although some clinical variables (e.g., urinary albumin, serum ALT, AST, creatinine, γ-glutamyl transpeptidase, lipoprotein A) appeared to have a high clinical sensitivity or specificity, the overall accuracy when considering both sensitivity and specificity was not satisfactory, indicating the need for a more accurate approach. For example, the AUC for urinary albumin was 0.51 (0.31–0.72; Table 2). Specifically, we built a multivariable model by combining our 5hmC-based panel and various clinical variables from the training set and evaluated the performance in the testing set. In the testing set, the AUC of weighted scores were larger than that of these clinical variables. And the AUC of these clinical variables combined with weighted scores also increased (Table 2). Notably, the multivariable model showed an independent indicator of our weighted score for detecting patients with vascular complications.

Table 1.

Sensitivity and specificity of traditional clinical variables.

Clinical parameter With or without complications
 Single or multiple complications
Sensitivity Specificity Sensitivity Specificity
Glucose 0.64 0.60 0.63 0.33
Hb A1c 0.54 0.33 0.56 0.50
ALT 0.12 0.73 0.09 0.81
AST 1.00 0.20 0.09 0.81
GGTa 0.11 0.82 0.09 0.87
ALP 0.02 0.91 0.00 0.33
LDL-C 0.26 0.55 0.30 0.80
HDL-C 0.31 0.82 0.33 0.73
Lp(a) 0.16 0.91 0.11 0.73
NEFA 0.21 0.73 0.18 0.73
CREA 0.24 0.92 0.21 0.69
UA 0.63 0.75 0.61 0.33
BUN 0.28 0.83 0.19 0.53
eGFR 0.70 0.42 0.71 0.31
u-Alb 0.82 0.25 0.85 0.25
Open in a new tab
a

GGT, gamma-glutamyl transpeptidase; ALP, alkaline phosphatase; LDL-C, LDL cholesterol; HDL-C, HDL cholesterol; Lp(a), lipoprotein A; NEFA, nonesterified fatty acid; CREA, creatinine; UA, uric acid; BUN, blood urea nitrogen; u-Alb, Urinary albumin.

Table 2.

Performance of the weighted scores and traditional clinical variables in the testing set.

With or without complications
 Single or multiple complications
AUC 95% CI AUC combined with WSa 95% CI AUC 95% CI AUC combined with WS 95% CI
WS 0.85 0.73–0.96 – – 0.84 0.68–0.99 – –
BMI 0.61 0.31–0.90 0.85 0.73–0.96 0.49 0.22–0.75 0.84 0.68–1.00
Smoking 0.46 0.21–0.71 0.82 0.68–0.96 0.59 0.34–0.85 0.85 0.69–1.00
Diabetic duration 0.67 0.41–0.93 0.86 0.75–0.98 0.58 0.35–0.82 0.84 0.68–0.99
DBP 0.53 0.28–0.77 0.86 0.75–0.98 0.70 0.47–0.93 0.84 0.69–0.99
SBP 0.70 0.40–1.00 0.84 0.71–0.97 0.70 0.46–0.93 0.83 0.67–0.99
Glucose 0.50 0.22–0.78 0.74 0.53–0.96 0.67 0.23–1.00 0.90 0.72–1.00
Hb A1c 0.52 0.20–0.84 0.86 0.73–0.98 0.57 0.33–0.82 0.84 0.68–0.99
ALT 0.62 0.35–0.90 0.83 0.70–0.97 0.57 0.31–0.83 0.82 0.64–0.99
AST 0.58 0.35–0.82 0.84 0.71–0.97 0.56 0.31–0.81 0.83 0.66–0.99
GGT 0.53 0.30–0.77 0.83 0.70–0.97 0.62 0.38–0.87 0.83 0.66–1.00
ALP 0.52 0.28–0.75 0.87 0.73–1.00 0.61 0.37–0.86 0.84 0.67–1.00
LDL-C 0.65 0.30–1.00 0.78 0.63–0.93 0.56 0.29–0.83 0.85 0.66–1.00
HDL-C 0.64 0.40–0.88 0.78 0.63–0.93 0.51 0.24–0.78 0.82 0.63–1.00
Lp(a) 0.54 0.31–0.78 0.79 0.64–0.94 0.46 0.20–0.73 0.82 0.64–1.00
NEFA 0.61 0.35–0.86 0.79 0.64–0.94 0.63 0.36–0.89 0.84 0.66–1.00
CREA 0.53 0.24–0.81 0.86 0.74–0.98 0.75 0.49–1.00 0.82 0.63–1.00
UA 0.54 0.25–0.82 0.86 0.74–0.98 0.52 0.27–0.77 0.79 0.60–0.98
BUN 0.49 0.22–0.75 0.86 0.74–0.98 0.60 0.34–0.87 0.80 0.62–0.99
eGFR 0.53 0.25–0.81 0.86 0.75–0.98 0.60 0.37–0.82 0.85 0.69–1.00
u-Alb 0.51 0.31–0.72 0.86 0.73–0.98 0.65 0.41–0.89 0.84 0.68–1.00
Open in a new tab
a

WS, weighted score; DBP, diastolic blood pressure; SBP, systolic blood pressure; GGT, gamma-glutamyl transpeptidase; ALP, Alkaline phosphatase; LDL-C, LDL cholesterol; HDL-C, HDL cholesterol; Lp(a), lipoprotein A; NEFA, nonesterified fatty acid; CREA, creatinine; UA, uric acid; BUN, blood urea nitrogen; eGFR, estimated glomerular filtration rate; u-Alb, urinary albumin.

PATHWAY ANALYSIS AND FUNCTIONAL EXPLORATION

Functional annotation analysis of the 135 candidate marker genes between patients with or without complications (see Table 2 in the online Data Supplement) suggested enrichment in certain canonical pathways. The pathways of the top 10% false discovery rate corrected P value and number of genes in each pathway ≥3 included signaling by insulin like growth factor 1 receptor (IGF1R), interleukin-10 (IL-10), and fibroblast growth factor receptor 2 (Fig. 4A). Of these pathways, signaling by IGF1R and IL-10 are known to be relevant to insulin resistance and inflammation, which play important roles in vascular biology and diabetes. The hubs of the Reactome functional interaction networks (Fig. 4B) showed that ubiquitin encoding genes [ubiquitin B (UBB),8 ubiquitin C (UBC), ubiquitin A-52 residue ribosomal protein fusion product 1 (UBA52)] were the central linkers of candidate genes [i.e., SMAD family member 3 (SMAD3), forkhead box D2 (FOXD2), DR1 associated protein 1 (DRAP1), multiciliate differentiation and DNA synthesis associated cell cycle protein (MCIDAS), motilin (MLN), PILR alpha associated neural protein (PIANP), and tubulin folding cofactor C (TBCC)]. Of these 135 candidate marker genes, a few genes [i.e., TNF receptor superfamily member 25 (TNFRSF25), P = 0.027, HR = 6.25; MLN, P = 0.0021, HR = 20.40; C-C motif chemokine ligand 3 like 1 (CCL3L1), P = 0.018, HR = 0.44] showed the trend of association with diabetic duration, suggesting the possibility of time-dependent interaction between 5hmC and diabetic vascular complications (see Table 4 in the online Data Supplement). Future studies based on longitudinal sampling could further establish the dynamic relationship between 5hmC changes and diabetic duration.

Fig. 4. Pathway analysis and function exploration.

Fig. 4.

Open in a new tab

(A), Pathways enriched in the 135 candidate marker genes with modified 5hmC between patients with and without complications are shown. (B), A functional interaction (FI) sub-network is constructed based on the measurement of betweenness centrality across linker genes (red) and candidate genes (black). The node size is proportional to the betweenness centrality calculated from the network using the Reactome database. Human genes: CCL3L3, C-C motif chemokine ligand 3 like 3; CXCL1, C-X-C motif chemokine ligand 1; FGF20, fibroblast growth factor 20; IGF1, insulin like growth factor 1; FGF3, fibroblast growth factor 3; FGFBP1, fibroblast growth factor binding protein 1; COL1A2, collagen type I alpha 2 chain; COL1A1, collagen type I alpha 1 chain; PTPN11, protein tyrosine phosphatase non-receptor type 11; TBCE, tubulin folding cofactor E; JUN, Jun proto-oncogene, AP-1 transcription factor subunit; CTCF, CCCTC-binding factor; HOXC8, homeobox C8; GNB1, G protein subunit beta 1; PILRA, paired immunoglobin like type 2 receptor alpha.

Further functional annotation analysis of the 159 candidate marker genes between patients with single or multiple complications (see Table 3 in the online Data Supplement) showed enrichment in such pathways as cell–cell junction, collagen formation, metabolism, and cellular responses to stress, which are related to the fundamental cell functions (see Table 3 in the online Data Supplement).

Discussion

Our primary aim was to evaluate whether 5hmC in cfDNA could distinguish patients with T2D who had vascular complications from patients without such complications. Previous studies have implicated 5hmC as a novel class of biomarkers for various cancers and hematological malignancies (20), but the value of 5hmC in cfDNA for detecting vascular complications in the context of diabetes has not yet been investigated.

Vascular complications contribute to about two thirds of deaths in diabetic patients (22). Mortality is increased 10-fold and 5-year survival is only 12.5% in diabetic patients with heart failure (30). The prevalence of retinopathy and nephropathy in diabetic population are 28.5% and 20%–40%, respectively (31). The major risk factors include smoking, hypertension, hyperglycemia, dyslipidemia, and albuminuria. Although diabetic duration is a risk factor, effectively controlling those major risk factors could prevent or slow the development of vascular events (22). Specifically, exposure of blood vessels to hyperglycemia, dyslipidemia, and inflammatory cytokines triggers thickening of the vascular basement membrane, induces dysfunctions of endothelial cells, and disturbs the communication between endothelial cells and pericytes, and further leads to diabetic vasculopathy and atherosclerosis (4). Some unstable atherosclerotic plaques are prone to rupture (32), and inflammation could cause plaque disruption (33). Both unstable plaques and apoptotic/necrotic vascular endothelial cells could release DNA into the bloodstream that might be the source of cfDNA in circulating blood. We reasoned that developing a clinically convenient, noninvasive approach utilizing cfDNA to effectively detect patients with T2D with or without vascular complications would greatly improve the clinical outcomes for these patients. For example, for patients with T2D who are detected at an early stage of vascular complications, various preventive treatments could be used to control the development of severe clinical outcomes.

Importantly, the 16-gene Detection Model 1 showed not only good performance in distinguishing patients with or without complications but also superiority over commonly used clinical variables including diabetic duration, BMI, and eGFR (Table 1–2). For example, compared to the clinical sensitivity (0.70), clinical specificity (0.42), and overall accuracy (AUC = 0.53) of eGFR (Table 1–2), an indicator for kidney function, the 16-gene panel for vascular complications showed significantly improved detection accuracy (testing set: AUC = 0.85; 95% CI, 0.73–0.96). Similarly, the 13-gene Detection Model 2 outperformed those clinical variables in terms of detection accuracy for distinguishing patients with single complications from those with multiple ones. For example, CREA showed the highest AUC among the tested clinical variables (testing set: AUC = 0.75; 95% CI, 0.49–1.00; Table 2). In comparison, the AUC of the 13-gene panel was 0.84 (95% CI, 0.68–0.99; Table 1–2). Notably, when combining the weighted scores and these clinical variables, the AUCs generally reflected the detection accuracy of the weighted scores alone, suggesting that the detection capacity of 5hmC marker panels was independent of those clinical variables (Table 2).

To our knowledge, the 5hmC-Seal is the only method available that allows highly sensitive mapping of genome-wide 5hmC in limited cfDNA materials. Our findings therefore not only revealed the detection capacity of the 5hmC markers in cfDNA for vascular complications but also supported the potential biological mechanisms for the development of vascular complications in patients with T2D. Specifically, pathway analysis of the 135 candidate marker genes with differentially modified 5hmC between patients with or without complications suggested enrichment in IGF1R and IL-10 signaling pathways (34, 35) (Fig. 4A), which are relevant to insulin resistance and inflammation, a pathophysiological hallmark of T2D and a risk factor for developing vascular complications (36). Insulin resistance induces multiple pathologies in the cardiovascular system, including endothelial dysfunction, acceleration of atherosclerosis, and poor collateral vascular formation or angiogenesis (36). Moreover, inflammation likely leads to T2D by causing insulin resistance, which is in turn intensified in the condition of hyperglycemia to induce diabetic complications (37). Furthermore, the 5hmC marker genes for vascular complications appeared to be functionally connected based on the Reactome network analysis (Fig. 4B). Notably, ubiquitin C, which was identified to be a hub gene among several 5hmC marker genes for vascular complications, is a highly conserved 76 amino-acid protein that plays vital roles in a remarkably diverse set of fundamental cellular processes. Maintenance of sufficient cellular ubiquitin levels in response to different metabolic conditions is critical for cellular function and survival (38).

There are limitations to our current study. First, although we demonstrated that 5hmC profiling in cfDNA had the potential to be an approach for distinguishing vascular complications in a small size of patients with T2D, more independent participants and validation in an independent cohort of sample set are necessary to refine and validate the marker panel. Second, our study design and patient features did not allow us to predict when a patient might develop related complications and establish the lead time for detection. Future studies using a longitudinal study design and series blood collection will be necessary to evaluate the predictive value of this approach for patients who do not yet have complications (i.e., preclinical). Third, our study was focused on Chinese patients, and thus, it was not possible to explore racial or population disparities, which would be an important research area for T2D, which is known to exhibit substantial racial or population differences in incidence rate and clinical outcomes.

In conclusion, taking advantage of our 5hmC-Seal technique, we demonstrated that 5hmCs in patient-derived cfDNA showed promise as a clinically convenient and noninvasive marker for T2D vascular complications, with the potential to complement other conventional clinical variables or risk factors for disease monitoring. Our findings warrant larger-scale studies using the 5hmC-Seal to further refine and validate the 5hmC markers for T2D-related vascular complications.

Acknowledgments

C. He thanks The University of Chicago Ludwig Center for partial support. C. He is a Howard Hughes Medical Institute Investigator.

7 Nonstandard abbreviations

T2D

Type 2 diabetes

5mC

5-methylcytosine

5hmC

5-hydroxymethylcytosine

cfDNA

cell-free DNA

eGFR

estimated glomerular filtration rate

ALT

alanine transaminase

AST

aspartate transaminase

Hb A1c

glycated hemoglobin

BMI

body mass index

HR

hazard ratio

AUC

area under the curve

IGF1R

insulin like growth factor 1 receptor

IL-10

interleukin-10.

8 Human Genes

UBB

ubiquitin B

UBC

ubiquitin C

UBA52

ubiquitin A-52 residue ribosomal protein fusion product 1

SMAD3

SMAD family member 3

FOXD2

forkhead box D2

DRAP1

DR1 associated protein 1

MCIDAS

multiciliate differentiation and DNA synthesis associated cell cycle protein

MLN

motilin

PIANP

PILR alpha associated neural protein

TBCC

tubulin folding cofactor C

TNFRSF25

TNF receptor superfamily member 25

CCL3L1

C-C motif chemokine ligand 3 like 1.

Footnotes

Author Contributions: All authors confirmed they have contributed to the intellectual content of this paper and have met the following 4 requirements: (a) significant contributions to the conception and design, acquisition of data, or analysis and interpretation of data; (b) drafting or revising the article for intellectual content; (c) final approval of the published article; and (d) agreement to be accountable for all aspects of the article thus ensuring that questions related to the accuracy or integrity of any part of the article are appropriately investigated and resolved.

Y. Yang, sample and clinical information collection, drafting manuscript; C. Zeng, statistical analysis, drafting manuscript; X. Lu, 5hmC-Seal profiling; Y. Song, 5hmC-Seal profiling; J. Nie, technical support; R. Ran, sample collection; Z. Zhang, statistical analysis; C. He, conceptualization, financial support; W. Zhang, conceptualization, analytical methodology, drafting manuscript, financial support; S.M. Liu, conceptualization, drafting manuscript, financial support. All authors approved the final manuscript.

Authors' Disclosures or Potential Conflicts of Interest: Upon manuscript submission, all authors completed the author disclosure form. Disclosures and/or potential conflicts of interest:

Employment or Leadership: X. Lu, Shanghai Epican Genetech Co. Ltd.; Y. Song, Shanghai Epican Genetech Co. Ltd.; C. He, a scientific founder of Accent Therapeutics, Inc.

Consultant or Advisory Role: C. He, Accent Therapeutics, Inc.; W. Zhang, Shanghai Epican Genetech Co., Ltd.

Stock Ownership: C. He, Accent Therapeutics, Inc., Shanghai Epican Genetech Co. Ltd.; W. Zhang, Shanghai Epican Genetech Co., Ltd.; X. Lu, Shanghai Epican Genetech Co. Ltd.; Y. Song, Shanghai Epican Genetech Co. Ltd.

Honoraria: None declared.

Research Funding: S.M. Liu, the National Natural Science Foundation of China (81972009, 81772276, 91753201, 81472023), and Health Commission of Hubei Province Scientific Research Project (WJ2019H005, WJ2019C002); W. Zhang, P30 CA060553 Career Development Fund from the National Institutes of Health; C. He, The University of Chicago Ludwig Center; Y. Yang, Program of Excellent Doctoral (Post Doctoral) of Zhongnan Hospital of Wuhan University (ZNYB2019013).

Expert Testimony: None declared.

Patents: The 5hmC-Seal technology was invented by C. He and was licensed by Shanghai Epican Genetech Co. Ltd. for clinical diagnosis and prognosis of human diseases from the University of Chicago, patent no. 8,741,567.

Role of Sponsor: The funding organizations played no role in the design of study, choice of enrolled patients, review and interpretation of data, or final approval of manuscript.

Contributor Information

Chang Zeng, Email: smliu@whu.edu.cn.

Zhou Zhang, Email: wei.zhang1@northwestern.edu.

Chuan He, Email: chuanhe@uchicago.edu.

References

  • 1. International Diabetes Federation. IDF diabetes atlas, 8th ed Brussels: International Diabetes Federation; 2017. [Google Scholar]
  • 2. Zimmet P, Alberti KG, Magliano DJ, Bennett PH. Diabetes mellitus statistics on prevalence and mortality: facts and fallacies. Nat Rev Endocrinol 2016;12:616–22. [DOI] [PubMed] [Google Scholar]
  • 3. Reddy MA, Zhang E, Natarajan R. Epigenetic mechanisms in diabetic complications and metabolic memory. Diabetologia 2015;58:443–55. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4. Wimmer RA, Leopoldi A, Aichinger M, Wick N, Hantusch B, Novatchkova M, et al. Human blood vessel organoids as a model of diabetic vasculopathy. Nature 2019;565:505–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5. Zheng Y, Ley SH, Hu FB. Global aetiology and epidemiology of type 2 diabetes mellitus and its complications. Nat Rev Endocrinol 2018;14:88–98. [DOI] [PubMed] [Google Scholar]
  • 6. Litwak L, Goh SY, Hussein Z, Malek R, Prusty V, Khamseh ME. Prevalence of diabetes complications in people with type 2 diabetes mellitus and its association with baseline characteristics in the multinational A1chieve study. Diabetol Metab Syndr 2013;5:57. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7. Tahiliani M, Koh KP, Shen Y, Pastor WA, Bandukwala H, Brudno Y, et al. Conversion of 5-methylcytosine to 5-hydroxymethylcytosine in mammalian DNA by MLL partner TET1. Science 2009;324:930–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8. Kriaucionis S, Heintz N. The nuclear DNA base 5-hydroxymethylcytosine is present in Purkinje neurons and the brain. Science 2009;324:929–30. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9. Fu Y, He C. Nucleic acid modifications with epigenetic significance. Curr Opin Chem Biol 2012;16:516–24. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10. Li W, Liu M. Distribution of 5-hydroxymethylcytosine in different human tissues. J Nucleic Acids 2011:870726, 2011. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11. Song CX, Yin S, Ma L, Wheeler A, Chen Y, Zhang Y, et al. 5-hydroxymethylcytosine signatures in cell-free DNA provide information about tumor types and stages. Cell Res 2017;27:1231–42. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12. Wang J, Tang J, Lai M, Zhang H. 5-hydroxy-methylcytosine and disease. Mutat Res Rev Mutat Res 2014;762:167–75. [DOI] [PubMed] [Google Scholar]
  • 13. Wu X, Zhang Y. TET-mediated active DNA demethylation: mechanism, function and beyond. Nat Rev Genet 2017;18:517–34. [DOI] [PubMed] [Google Scholar]
  • 14. Liang J, Yang F, Zhao L, Bi C, Cai B. Physiological and pathological implications of 5-hydroxymethylcytosine in diseases. Oncotarget 2016;7:48813–31. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15. Cai J, Chen L, Zhang Z, Zhang X, Lu X, Liu W, et al. Genome-wide mapping of 5-hydroxymethylcytosines in circulating cell-free DNA as a non-invasive approach for early detection of hepatocellular carcinoma. Gut. 2019July29 pii: gutjnl-2019-318882. doi: 10.1136/gutjnl-2019-318882. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16. Kowluru RA, Shan Y, Mishra M. Dynamic DNA methylation of matrix metalloproteinase-9 in the development of diabetic retinopathy. Lab Invest 2016;96:1040–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17. Pinzon-Cortes JA, Perna-Chaux A, Rojas-Villamizar NS, Diaz-Basabe A, Polania-Villanueva DC, Jacome MF, et al. Effect of diabetes status and hyperglycemia on global DNA methylation and hydroxymethylation. Endocr Connect 2017;6:708–25. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18. Gravina S, Sedivy JM, Vijg J. The dark side of circulating nucleic acids. Aging cell 2016;15:398–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19. Li W, Zhang X, Lu X, You L, Song Y, Luo Z, et al. 5-hydroxymethylcytosine signatures in circulating cell-free DNA as diagnostic biomarkers for human cancers. Cell Res 2017;27:1243–57. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20. Zeng C, Stroup EK, Zhang Z, Chiu BC, Zhang W. Towards precision medicine: advances in 5-hydroxy-methylcytosine cancer biomarker discovery in liquid biopsy. Cancer Commun 2019;39:12. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21. Song CX, Szulwach KE, Fu Y, Dai Q, Yi C, Li X, et al. Selective chemical labeling reveals the genome-wide distribution of 5-hydroxymethylcytosine. Nat Biotechnol 2011;29:68–72. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22. American Diabetes Association. Standards of medical care in diabetes-2017: summary of revisions. Diabetes Care 2017;40(Suppl 1):S4–5. [DOI] [PubMed] [Google Scholar]
  • 23. American Diabetes Association. 10. Microvascular complications and foot care. Diabetes Care 2017;40(Suppl 1):S88–98. [DOI] [PubMed] [Google Scholar]
  • 24. Cheung N, Mitchell P, Wong TY. Diabetic retinopathy. Lancet 2010;376:124–36. [DOI] [PubMed] [Google Scholar]
  • 25. Spanopoulos D, Okhai H, Zaccardi F, Tebboth A, Barrett B, Busse M, et al. Temporal variation of renal function in people with type 2 diabetes mellitus: a retrospective UK clinical practice research datalink cohort study. Diabetes Obes Metab 2019;21:1817–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26. Han D, Lu X, Shih AH, Nie J, You Q, Xu MM, et al. A highly sensitive and robust method for genome-wide 5hmc profiling of rare cell populations. Mol Cell 2016;63:711–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27. Parikh R, Mathai A, Parikh S, Chandra Sekhar G, Thomas R. Understanding and using sensitivity, specificity and predictive values. Indian J Ophthalmol 2008;56:45–50. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28. Fabregat A, Jupe S, Matthews L, Sidiropoulos K, Gillespie M, Garapati P, et al. The Reactome pathway Knowledgebase. Nucleic Acids Res 2018;46(D1):D649–55. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29. Cox DR. Regression models and life-tables. J R Stat Soc Series B Stat Methodol 1972;34:187–202. [Google Scholar]
  • 30. Hippisley-Cox J, Coupland C. Diabetes treatments and risk of heart failure, cardiovascular disease, and all cause mortality: cohort study in primary care. BMJ 2016;354:i3477. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31. American Diabetes Association. 11. Microvascular complications and foot care: standards of medical care in diabetes-2019. Diabetes Care 2019;42(Suppl 1):S124–38. [DOI] [PubMed] [Google Scholar]
  • 32. Cubbon RM, Kearney MT, Wheatcroft SB. Endothelial IGF-1 receptor signalling in diabetes and insulin resistance. Trends Endocrinol Metab 2016;27:96–104. [DOI] [PubMed] [Google Scholar]
  • 33. Hai Z, Zuo W. Aberrant DNA methylation in the pathogenesis of atherosclerosis. Clin Chim Acta 2016;456:69–74. [DOI] [PubMed] [Google Scholar]
  • 34. Riddy DM, Delerive P, Summers RJ, Sexton PM, Langmead CJ. G protein-coupled receptors targeting insulin resistance, obesity, and type 2 diabetes mellitus. Pharmacol Rev 2018;70:39–67. [DOI] [PubMed] [Google Scholar]
  • 35. Ouyang W, Rutz S, Crellin NK, Valdez PA, Hymowitz SG. Regulation and functions of the IL-10 family of cytokines in inflammation and disease. Annu Rev Immunol 2011;29:71–109. [DOI] [PubMed] [Google Scholar]
  • 36. King GL, Park K, Li Q. Selective insulin resistance and the development of cardiovascular diseases in diabetes: the 2015 Edwin Bierman Award Lecture. Diabetes 2016;65:1462–71. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37. Lontchi-Yimagou E, Sobngwi E, Matsha TE, Kengne AP. Diabetes mellitus and inflammation. Curr Diab Rep 2013;13:435–44. [DOI] [PubMed] [Google Scholar]
  • 38. Ryu KY, Maehr R, Gilchrist CA, Long MA, Bouley DM, Mueller B, et al. The mouse polyubiquitin gene UbC is essential for fetal liver development, cell-cycle progression and stress tolerance. EMBO J 2007;26:2693–706. [DOI] [PMC free article] [PubMed] [Google Scholar]

Articles from Clinical Chemistry are provided here courtesy of Oxford University Press

RESOURCES