Department of Environmental Genomics, Jiangsu Key Laboratory of Cancer Biomarkers, Prevention and Treatment, Collaborative Innovation Center for Cancer Personalized Medicine, School of Public Health, Nanjing Medical University, Nanjing, Jiangsu 211166, China
2.
Department of Genetic Toxicology, the Key Laboratory of Modern Toxicology of Ministry of Education, Center for Global Health, School of Public Health, Nanjing Medical University, Nanjing, Jiangsu 211166, China
3.
Department of Laboratory Medicine, Zhenjiang Medical School of Nanjing Medical University, Zhenjiang, Jiangsu 212002, China
4.
Department of Endocrinology, Zhenjiang Medical School of Nanjing Medical University, Zhenjiang, Jiangsu 212002, China
5.
Department of Oncology, Zhenjiang Medical School of Nanjing Medical University, Zhenjiang, Jiangsu 212002, China
6.
Department of Laboratory Medicine, the Affiliated Hospital of Jiangsu University, Zhenjiang, Jiangsu 212008, China
Yingzhao Liu, Department of Endocrinology, Zhenjiang Medical School of Nanjing Medical University, 8 Dianli Road, Zhenjiang, Jiangsu 212002, China. E-mail: zjliuyingzhao@126.com
Zhengdong Zhang, Department of Environmental Genomics, School of Public Health, Nanjing Medical University, 101 Longmian Avenue, Nanjing, Jiangsu 211166, China. E-mail: drzdzhang@njmu.edu.cn
Most papillary thyroid carcinoma (PTC) patients have a good prognosis. However, lymph node metastasis (LNM), the most common manifestation of disease progression, is frequently associated with a poor prognosis. Nevertheless, few studies have focused on the underlying mechanisms of LNM. In the current study, we aimed to investigate the potential role of exosomal circRNAs that contribute to LNM in PTC. We identified 9000 differentially expressed exosomal circRNAs in PTC patients with LNM, including 684 upregulated and 2193 downregulated circRNAs. Functional enrichment analysis revealed that these differentially expressed circRNAs were primarily involved in a variety of molecular and signaling pathways correlated with PTC progression and LNM. Through bioinformatics analysis, we identified 14 circRNA-miRNA-mRNA networks related to LNM-associated signaling pathways in PTC. Moreover, both circTACC2-miR-7-EGFR and circBIRC6-miR-24-3p-BCL2L11 axes were verified for their potential involvement in PTC with LNM. Additionally, we identified four upregulated circRNA-related hub genes and eight hub genes correlated with downregulated circRNAs, some of which were validated as being potentially involved in LNM in PTC. Collectively, our findings provide a novel framework for an in-depth investigation of the function of dysregulated exosomal circRNAs and their potential as biomarkers in PTC patients with LNM.
Gestational diabetes mellitus (GDM) complicates approximately 14% of pregnancies globally, with a prevalence of 21% in Asia[1]. Women with a history of GDM face a seven to 10-fold increased risk of developing type 2 diabetes mellitus (T2D) and a two to three-fold higher risk of cardiovascular disease, compared with those without GDM[2-4]. Current guidelines (American Diabetes Association, 2024) recommend that all GDM women undergo an oral glucose tolerance test (OGTT) at four to 12 weeks postpartum[5]. Intensive lifestyle modifications and metformin treatment have demonstrated 35%–40% risk reductions in diabetes progression risk post-GDM versus placebo[6]. Thus, identifying prediabetes early in these women is crucial for enabling interventions to delay or prevent the progression to diabetes.
However, the compliance of global postpartum OGTT averages only 33%, ranging between 9% and 71%[7]. Diabetes was reported in 2.8%–58% of those women based on the length of follow-up, while theprevalence of prediabetes, which is an intermediate stage between normal glucose regulation and diabetes, was 3.9%–50.9% in early postpartum[8].OGTT has an accuracy rate of only 60%–70% to predict future T2D[9]. Factors such as the inconvenience and inaccuracy of the OGTT further contribute to the low follow-up rates. Thus, there is a pressing need to develop more convenient and precise reclassification methods for postpartum assessment because of currently limited current options.
Metabolomics offers a valuable approach to identifying metabolic changes associated with GDM and T2D. Several studies identified that circulating metabolites in the early postpartum period might predict of T2D in women with a history of GDM[9-11]. However, most of these studies have focused on fasting plasma or serum and have rarely addressed prediabetes, which is more common than T2D among women with previous GDM in the early postpartum period.
Prediabetes is a heterogeneous glucose metabolism disorder characterized by impaired fasting glucose (IFG) and impaired glucose tolerance (IGT), assessed by using a 2-hour post-load glucose test. It is important to note that plasma/serum circulating metabolites are significantly influenced by diet and glucose load. Thus, we measured numerous metabolites in stored frozen plasma samples collected at both six- week to six-month postpartum from women with a history of GDM, using a 2-hour 75 g OGTT, at both fasting and 2-hour post-load time points.
Our primary aim was to gain insights into the underlying pathophysiology of the transition from GDM to prediabetes in the early postpartum period through a metabolic approach under fasting and glucose load status, respectively. A secondary goal was to identify metabolic biomarkers of prediabetes for building a more convenient and accurate screening method in the future, to further strategize cost-effective post-GDM monitoring.
Materials and methods
Study design
We conducted a cross-sectional study of women with a history of GDM based on the 2-hour 75g OGTT during pregnancy. Women who delivered a singleton, live-born infant at the First Affiliated Hospital of Nanjing Medical University between April 2021 and May 2022 were recruited. Women with multiple gestations, pre-gestational diabetes, serious health conditions, or using medications affecting glucose tolerance were excluded from the study. Eligible women with a history of GDM were enrolled in the study and underwent a 2-hour 75g OGTT between six-week and six-month postpartum. As a result, 40 women had normal postpartum glucose tolerance (the control group), and 42 women were diagnosed with postpartum prediabetes (the prediabetes group), including seven with IFG, 31 with IGT, and four with both IFG and IGT. Diagnosis criteria for GDM and prediabetes were based on the criteria of the International Association of Diabetes and Pregnancy Study Groups (IADPS) and the American Diabetes Association (ADA), respectively[12]. Given the poor concordance between the OGTT and glycosylated hemoglobin for diagnosing diabetes and prediabetes following GDM[13-14], we only used the OGTT results as the diagnostic criteria for prediabetes in the current study.
Clinical data were collected from both patient visits and electronic medical records, including age, height, family history of diabetes, antepartum and postpartum laboratory results, pre-pregnancy weight, pre-delivery weight, postpartum weight, blood pressure, dates of diabetes diagnosis, and other clinical outcomes.
The area under the curve (AUC) of the OGTT time-blood glucose was calculated using the trapezoidal rule. Postpartum weight retention (PWR) was defined as the difference between postpartum weight and pre-pregnancy weight, while postpartum weight loss (PWL) was defined as the difference between postpartum weight and pre-delivery weight.
The current study protocol was approved by the Institutional Review Board Research Ethics Committee (REC) of the First Affiliated Hospital of Nanjing Medical University (approval number: 2023-SR-579). Before the enrollment, we obtained a written informed consent from each subject.
Plasma sample collection and preparation
Blood samples were collected from each participant at both fasting and 2-hour post-load time points during the 75g OGTT. The samples were centrifuged (1000g, 10 min, 4 ℃) and then stored at −80 ℃ for subsequent metabolomics analysis.
Metabolomics methods
Sample preparation for ultra-high performance liquid chromatography-quadrupole time of flight mass spectrometry (UHPLC-Q-TOF-MS/MS)
The pooled plasma, mixed with anequal aliquot of plasma sample from the control group and the prediabetes group, was used as quality control samples. Both plasma and quality control samples were mixed with the four times volume of pre-cooled methanol, vortexed for 30 s, frozen at −20 ℃ for 1 h, and centrifuged at 16000g for 15 min at 4 ℃. The supernatant was lyophilized to dryness, and then dissolved in an acetonitrile-water solution (1:1, v/v) to vortex and centrifuge again. Subsequently, 80 µL of the supernatant was transferred for UHPLC-Q-TOF-MS/MS analysis (AB Sciex, Framingham, MA, USA). The quality control samples were included in every 10 samples to guarantee instrument stability. The solvent in which the samples were dissolved was used as a blank to subtract the solvent peaks and eliminate systemic impurity signals. The starting and ending blank samples of each batch were examined to find any carryover or contamination.
UHPLC-Q-TOF-MS/MS analysis and data pre-processing
Chromatographic analysis was performed on a SCIEX ExionLC UHPLC system (AB Sciex, Framingham, MA, USA) using an Atlantis™ Premier BEH C18 AX column (2.1 × 100 mm, 1.7 μm) (Waters, Milford, MA, USA) with a guard column (VanGuard™ FIT, Waters, Milford, MA, USA) for chromatographic separation. The column temperature was 40 ℃ and the injection volume was 5 μL. Gradient elution was carried out at a flow rate of 0.4 mL/min with 0.1% formic acid (Merck, Germany) in water (solvent A) and acetonitrile (Merck, Germany) (solvent B): 0–1 min, 1% B; 1–10 min, 1%–99% B; 10–13 min, 99% B; 13–14 min, 99%–1% B; and 14–17 min, 1% B.
Data acquisition in both positive and negative ion modes was performed using a TripleTOFTM5600+ system (AB Sciex, Framingham, MA, USA) equipped with a DuoSpray Source (electrospray ionization, ESI) (AB Sciex, Framingham, MA, USA). The ESI conditions were set as follows: ion source gas 1 at 55 psi, ion source gas 2 at 55 psi, curtain gas at 35 psi, and spray voltage at 5.5 kV (+)/−4.5 kV (−). The time-of-flight mass spectrometry information-dependent acquisition mass spectrometry (TOF MS-IDA-MS/MS) acquisition included a TOF MS scan and a product ion scan based on information-dependent acquisition (IDA). The TOF MS scan range was 50–1000 m/z, and the TOF IDA-MS/MS (product ion) scan range was 40–1000 m/z. The IDA mode settings were: declustering potential of 80 V (+)/−80 V (−), collision energy of 35 [± 15] V (+)/−35 [± 15] V (−). Dynamic background subtraction was applied to match the IDA tests for UHPLC-Q-TOF-MS/MS. A calibration delivery system was used for automatic calibration of TOF MS and TOF MS/MS every five samples.
Data acquisition was performed using Analyst TF 1.8.1 (AB Sciex, Framingham, MA, USA), and then the raw data (.wiff files) were converted into the .mzML format using ProteoWizard software (Vanderbilt University, Nashville, TN, USA). The XCMS software (Scripps Research Institute, La Jolla, USA) was used for peak alignment, retention time correction, and peak area extraction.
Statistical analysis
The significance statistics were performed with SPSS Statistics version 26 (IBM, Armonk, USA), including the Chi-square test for categorical variables (n and %), the Student’s t-test for normally distributed continuous variables (mean and standard deviation), and the Mann-Whitney test for non-normally distributed variables (median and interquartile ranges).
The univariable and multidimensional statistical analyses were conducted using the “R” language to detect significantly differential metabolites between the control and prediabetes groups at the fasting and 2-hour post-load time points. The unsupervised principal component analysis (PCA) and supervised partial least squares discriminant analysis (PLS-DA) were used to visualize group patterns and calculate variable importance in projection (VIP) values of the metabolites. Metabolic pathway analysis and metabolite set enrichment analysis were carried out on the MetaboAnalyst 6.0 platform (https://www.metaboanalyst.ca/MetaboAnalyst/home.xhtml; accessed on July 18, 2024). A network diagram illustrating the interactions between metabolites and genes was conducted using Cytoscape version 3.10.2 software. We performed the plotting of the receiver operator characteristic (ROC) curves and compared the AUCs using the SigmaPlot 14.0 software package. Differences with a P < 0.05 were considered statistically significant.
Results
Demographic and clinical characteristics
The demographic characteristics of the subjects in both groups are summarized in Table 1. The subjects were older in the prediabetes group than in the control group (P = 0.010). There were no significant differences in terms of parity, family history of diabetes, feeding patterns, postpartum weight, PWL, PWR, postpartum blood pressure, postpartum waist, and postpartum hip circumference between the prediabetes group and the control group.
Table
1.
Postpartum demographic characteristics of the subjects
Characteristics
Control (n=40)
Prediabetes (n=42)
P-value
Age (years, median [Q1, Q3])
31 (29–33)
33 (30–35)
0.010*
Multipara (n [%])
8 (20.0%)
16 (38.1%)
0.072
Family history of diabetes (n [%])
17 (42.5%)
17 (40.5%)
0.852
Exclusive breastfeeding (n [%])
20 (50%)
16 (38.1%)
0.278
Weight (kg, mean ± SD)
60.4±9.0
61.4±7.2
0.594
PWL (kg, mean ± SD)
8.3±3.0
8.6±2.8
0.572
PWR (kg, mean ± SD)
3.5±5.0
3.4±3.8
0.984
BMI (kg/m2, mean ± SD)
23.0±3.1
23.4±2.5
0.612
SBP (mmHg, mean ± SD)
112±10
112±12
0.959
DBP (mmHg, mean ± SD)
70±7
71±9
0.546
Waist circumference (cm, mean ± SD)
82.1±8.4
82.9±9.2
0.698
Hip circumference (cm, mean ± SD)
94.7±5.9
96.0±6.1
0.308
Waist-hip ratio (mean ± SD)
0.87±0.05
0.86±0.06
0.718
Data are expressed as means ± standard deviation (SD), medians (interquartile ranges), or numbers (percentages). Differences between the two groups were tested by the Chi-square test for categorical variables, the Student's t-test for normally distributed continuous variables, and the Mann-Whitney test for non-normally distributed variables. *P < 0.05. Abbreviations: BMI, body mass index; SBP, systolic blood pressure; DBP, diastolic blood pressure.
As shown in Fig. 1, at the fasting time point, there were no significant differences in blood glucose levels during pregnancy and postpartum, as well as postpartum insulin and C-peptide levels, between the prediabetes group and the control group (Fig. 1A, 1B, 1C, and 1D). At one hour after oral glucose administration, only the levels of blood glucose postpartum was significantly increased in the prediabetes group than in the control group, and there was no difference in postpartum insulin, C-peptide and blood glucose during pregnancy (Fig. 1E, 1F, 1G, and 1H). However, two hours after oral glucose administration, the levels of blood glucose, insulin, and C-peptide during both pregnancy and postpartum were significantly elevated in the prediabetes group than in the control group (Fig. 1I, 1J, 1K, and 1L).
Figure
1.
Box plots of comparing the OGTT results and clinical characteristics during pregnancy and postpartum between the prediabetes group (n=42) and the control group (n=40).
The top and bottom of the box plot represents the 25th and 75th percentile range. The line across the box shows the median and the top and bottom bars show the maximum and minimum values. (A) Glucose levels during the pregnancy OGTT, (B) glucose, (C) insulin and (D) C-peptide levels during the postpartum OGTT at the fasting time point. (E) Glucose levels during the pregnancy OGTT, (F) glucose, (G) insulin and (H) C-peptide levels during the postpartum OGTT at the 1-hour post-load time points. (I) Glucose levels during the pregnancy OGTT, (J) glucose, (K) insulin and (L) C-peptide levels during the postpartum OGTT at the 2-hour post-load time points. The AUCs of the time-blood glucose during the pregnancy OGTT (M) and postpartum OGTT (N). (O) Postpartum Matsuda index. (P) IGI60. (Q) HOMA-β. (R) HOMA-IR. (S) HbA1c. Data are presented as medians (interquartile ranges).P-values were determined by the Mann-Whitney test. *P < 0.05, **P < 0.01, and ***P < 0.001. Abbreviations: ap, antepartum; pp, postpartum; GLU, glucose; INS, insulin. GLU0, GLU60, and GLU120 indicate glucose levels at fasting, 1 hour, and 2 hours after oral glucose administration, respectively. INS and C-peptide levels at the same time points are denoted as INS0, INS60, INS120, and C-peptide0, C-peptide60, C-peptide120.
The AUCs of the OGTT time-blood glucose curve were significantly greater in the prediabetes group than in the control group, during both pregnancy and postpartum periods (Fig. 1M and 1N).
Compared with those in the control group, the subjects in the prediabetes group showed both a lower Matsuda index (P = 0.025) and a lower insulinogenic index 60 (P = 0.012) (Fig. 1O and 1P). There were no significant differences in the homeostatic model assessment for beta cell function (Fig. 1Q), the homeostatic model assessment for insulin resistance (Fig. 1R), and glycosylated hemoglobin (Fig. 1S) between the two groups.
Metabolome spectrum for UHPLC-Q-TOF-MS/MS
To explore metabolic differences between the two groups, we performed a comprehensive metabolomic analysis of the plasma from the subjects via UHPLC-Q-TOF-MS/MS. The overall distribution trends among all samples were observed using the PCA analysis (Fig. 2A and 2B). The PCA scores for the negative ion mode showed a significant separation between the two groups at different time points, while the positive ion mode displayed an insignificant trend. We performed PLS-DA to further differentiate the metabolite features and identify potential marker metabolites. The PLS-DA score plots showed well-distinguished differences between the groups at both time points in negative ion mode, with a partial overlap in positive ion mode at the fasting time point (Fig. 2C and 2D).
Figure
2.
The PCA and PLS-DA score plot of each group and metabolites with differences between the control (n=40) and prediabetes (n=42) groups at fasting and 2-hour post-load time points.
(A–D): The PCA and PLS-DA score plot for negative (A and C) and positive (B and D) ion modes, respectively. Each point represented each sample. Different groups were in different colors, with circles representing the 95% confidence interval. (E–F): Volcano plots of metabolites between groups at the fasting (E) and 2-hour post-load (F) time points. The increased and decreased metabolites were represented as red or blue scales, respectively. (G–H): Venn diagrams of the overlap metabolites detected between the two groups at separate time points, the increased metabolites are shown in G and the decreased metabolites are shown in H. The significance was determined using P-values adjusted by the Benjamini-Hochberg method for either Student’s t-test or Mann-Whitney test. *P < 0.05, ** P < 0.01, and *** P < 0.001. Abbreviations: PCA, principal component analysis; PLS-DA, partial least squares discriminant analysis.
Using the VIP value from the PLS-DA model (threshold > 1), a total of 164 metabolites exhibited differences between the two groups at separate time points (Supplementary Fig. 1). Metabolites with qualitatively significant differences were selected to construct hierarchical cluster plots for samples within each group, which were then used to evaluate the stability of the expression patterns of the selected target metabolites. Metabolites within similar clusters were presumed to share expression patterns, potentially indicating the presence of corresponding reaction steps in the metabolic pathways.
Taking into account the P-value from the Student's t-test (P < 0.05), at the fasting time point, the levels of 55 metabolites showed significant differences between the control and the prediabetes groups (Table 2 and Fig. 2E). Specifically, the levels of 20 metabolites, such as isocitrate, gamma-aminobutyric acid, and L-glutamic acid, were significantly increased, but other 35 metabolites, such as cholic acid, β-alanine, and L-proline, were significantly decreased in the prediabetes group than in the control group. At the 2-hour post-load time point, 58 metabolites exhibited significant differences in the expression in the prediabetes group, compared with the control group (Table 2 and Fig. 2F), among which, the levels of 39 metabolites, such as L-leucine, pyruvate, and chenodeoxycholic acid, were significantly increased but other 19 metabolites, such as cholic acid, L-glutamine, and fumarate, were significantly decreased.
Table
2.
Metabolites detected between the groups at different time points
Metabolites
Fasting
Metabolites
Two-hour post-load
log2(FC)
P-value
log2(FC)
P-value
(E)-cinnamaldehyde
0.64
**
(R)-carnitine
0.36
**
(S)-allantoin
0.15
*
1,3,7-Trimethyluric acid
−0.22
*
(S)-lactatea
−0.35
***
11-deoxyprostaglandin E1
−0.55
*
1,3,7-Trimethyluric acid
−0.24
**
2(5H)-Furanone
0.21
***
10-Hydroxydecanoic acid
0.25
*
2-Butynedioic acid
0.03
*
11-deoxyprostaglandin E1
−0.35
**
2-deoxy-D-glucosea
−0.31
***
2-aminoacrylic acid
−0.17
**
2-oxoadipic acid
0.15
***
2-deoxy-D-glucosea
−0.16
**
5,6-dihydroxyindole
−0.35
***
3,4-dihydroxymandelic acid
0.09
*
5-acetamido-6-formamido-3-methyluracil
0.49
***
4-hydroxyphenylacetic acid
0.12
*
7-methylxanthine
0.23
*
7,9-dihydro-1H-purine-2,6,8(3H)-trione
−0.28
**
acetoin
0.11
*
7-methylxanthine
−0.25
*
acrylate
0.21
***
allantoin
−0.43
**
allantoin
−0.4
*
alpha-Ketocaproic acid
0.21
*
alpha-D-galactose
0.22
***
arachidonic acid
0.44
***
benzene-1,2,4-triol
0.23
***
benzoic acid
0.17
**
butane-2,3-dione
0.25
***
beta-alanine
−0.45
**
chenodeoxycholic acidc
0.24
*
cholic acidc
−0.32
*
cholic acidc
−0.27
***
D-aspartic acid
−0.49
***
corticosterone
0.24
**
D-glucono-1,5-lactone
−0.48
***
cortisone
0.25
*
dodecanedioic acid
0.42
*
decanoate
0.38
*
dodecanoate
−0.56
*
decanoic acid
−0.28
***
D-Pyroglutamic acidb
0.19
**
D-fructofuranose 6-phosphate
0.14
**
D-tagatopyranose
−0.37
**
D-glucuronate
0.16
*
dUMP
−0.63
***
D-Pyroglutamic acidb
0.07
*
equol
0.34
*
dUMP
−0.25
*
estrone
0.6
*
fumaratea
−0.15
***
fumaratea
−0.07
*
galactose
0.51
***
fumaric acid
−0.24
**
glutaric acid
0.17
***
gamma-aminobutyric acidb
0.09
**
glycine betaine
0.1
*
homovanillic acid
−0.06
**
glycocholic acidbc
−0.2
*
hyocholic acidc
0.25
*
glycolate
0.28
***
hypoxanthine
−0.66
***
glycolic acid
−0.46
***
isocitric acida
0.23
**
glyoxylic acid
0.26
***
L-glutamic acidb
0.26
**
hyocholic acidc
0.72
*
linoleate
0.23
*
hyodeoxycholic acid
0.12
*
L-proline
−0.23
***
indol-3-ylacetaldehyde
0.61
***
L-Rhamnose
−0.11
*
indole-3-acetic acid
0.35
***
L-threonic acid
−0.44
***
L-arabinitol
0.13
***
L-tyrosine
−0.35
**
L-glutamineb
−0.08
*
maltose
−0.44
***
L-leucine
0.28
**
methylmalonic acid
−0.51
***
L-Rhamnose
−0.44
***
N-acetyl-L-alanine
0.22
*
L-sorbose
0.23
***
N-acetylneuraminic acid
−0.28
**
Methionine sulfoxide
0.13
*
N-benzoylglycine
−0.23
*
methylmalonate
0.17
***
N-cinnamoylglycine
−0.08
**
myo-inositol
0.3
***
octadecanoic acid
0.21
**
N-acetyl-L-proline
0.24
***
octanoic acid
0.17
*
N-acetylneuraminic acid
−0.38
*
oxomalonic acid
−0.15
*
N-cinnamoylglycine
−0.18
***
p-hydroxyhippuric acid
−0.41
***
octanoic acid
0.39
**
Purine
−0.49
*
oxomalonic acid
−0.19
***
pyridoxal
−0.51
*
pyruvatea
0.26
***
sucrosea
−0.35
***
quinolin-2-ol
−0.27
***
uric acid
−0.51
***
S-sulfo-L-cysteine
0.29
*
xylitol
−0.21
*
theobromine
0.32
***
thymidine
−0.51
***
tyramine sulfate
0.15
**
xylitol
−0.29
**
aMetabolites associated with energy metabolism. bMetabolites associated with glutamate metabolism. cMetabolites associated with bile acid metabolism.Concentration changes of metabolites are indicated by a color gradient, with red for increased and blue for decreased levels, based on the log-transformed fold change (log2[FC]). The significance was determined using P-values adjusted by the Benjamini-Hochberg method for either Student's t-test or Mann-Whitney test. *P < 0.05, **P < 0.01, and ***P < 0.001.
Notably, compared with the control group, 15 metabolites in the prediabetes group showed consistent change trends at both the fasting and 2-hour post-load time points, which included three consistently increased metabolites (D-pyroglutamic acid hyocholic acid, and octanoic acid) and 12 consistently decreased metabolites (1,3,7-trimethyluric acid, 11-deoxy prostaglandin E1, 2-deoxy-D-glucose, allantoin, cholic acid, dUMP, fumarate, L-rhamnose, N-acetylneuraminic, N-cinnamoylglycine, oxomalonic acid, and xylitol) (Fig. 2G and 2H).
Metabolic pathway analysis
Based on the results of PLS-DA, we inputted differential metabolites with VIP > 1 into the MetaboAnalyst 6.0 platform for metabolic pathway analysis and metabolite set enrichment analysis between the two groups. The Fig. 3A and 3B displayed the main enriched metabolic pathways. The metabolite set enrichment analysis revealed significant enrichment in amino acid metabolism pathways and energy metabolism pathways, including alanine, aspartate, and glutamate metabolism, and other metabolism pathways that were closely related to a disorder of glucose metabolism. Additionally, we employed the Cytoscape 3.10.2 software to construct a network of interactions between the differential metabolites and their upstream or downstream genes. A network diagram illustrating these interactions between the metabolites and genes was also created (Fig. 3C and 3D). These metabolites provided a more sensitive reflection of metabolic abnormalities in the prediabetes group than in the control group, which were involved in various pathways, including those shown in Fig. 3E.
Figure
3.
Metabolic profiling of pathway analysis, post-load interaction network, and the GDM to prediabetes progression pathways.
(A): Bubble graph of metabolic pathway analysis. (a) glyoxylate and dicarboxylate metabolism; (b) alanine, aspartate, and glutamate metabolism; (c) valine, leucine, and isoleucine biosynthesis; (d) glycine, serine, and threonine metabolism; (e) phenylalanine, tyrosine, and tryptophan biosynthesis; (f) citrate cycle; (g) tyrosine metabolism; (h) arginine and proline metabolism; (i) phenylalanine metabolism; (j) butanoate metabolism, (k) pyruvate metabolism. (B): Bar graph of metabolite set enrichment analysis. (C–D): Network of interactions between metabolites and genes at the 2-hour post-load time point. C shows the valine, leucine and isoleucine metabolism. D shows the bile acid biosynthesis. Hexagonal nodes and circular nodes represent metabolites and genes, respectively. Red-colored hexagons indicate differential metabolites with variable importance in projection (VIP) >1. The largest hexagons denote an increase, while the smallest denote a decrease. Those with green borders indicate a comparison between the two groups with P < 0.05. Pink-colored hexagons represent metabolites associated with the pathways. (E): Integrated metabolic pathways correlated with progression from gestational diabetes mellitus to prediabetes. Abbreviations: VIP, variable importance in projection.
Leucine is involved in the metabolism of branched-chain amino acids (BCAAs), and the elevated levels of BCAAs have been shown to be closely associated with the development of insulin resistance and type 2 diabetes[15]. The accumulation of pyruvate, which is involved in pyruvate metabolism, may trigger inflammation, thereby impairing the insulin signaling pathway. Fumarate and isocitrate are key intermediates of the tricarboxylic acid (TCA) cycle, and the impaired TCA cycle flux in insulin-resistant skeletal muscle is one of the characteristics of the diabetic phenotype.
Prediction for the progression from GDM to prediabetes
Using the ROC curve, we assessed the predictive abilities of three clinical markers during pregnancy for the progression from GDM to prediabetes, including age, 2-hour post-load glucose levels, and the AUC of glucose levels during the pregnancy OGTT (Fig. 4A). The AUCs of three clinical markers ranged from 0.65 to 0.68. The model combined these three clinical markers yielded an AUC of 0.71 (95% CI, 0.60–0.82), with a sensitivity of 0.46 and a specificity of 0.90 at the optimal cutoff. However, this increase was not statistically significant.
Figure
4.
ROC curves.
A: ROC curves of predictive models for the progression from GDM to prediabetes generated by age, 2-hour glucose levels, and the AUC of glucose during the pregnancy OGTT. B–C: ROC curves for the discriminative power of the panel of 15 metabolites for distinguishing between postpartum prediabetes and normal glucose tolerance at the fasting (B) and 2-hour post-load (C) time points. Abbreviations: ap, antepartum; GLU, glucose; AUC, area under the curve.
Given that the 15 metabolites in prediabetes showed consistent change trends before and after glucose loading, we also assessed the discriminative power of the panel of 15 metabolites for distinguishing between postpartum prediabetes and normal glucose tolerance at the fasting and 2-hour post-load time point (Fig. 4B and 4C). We achieved the AUC of 0.98 (95% CI, 0.94–1.00), sensitivity of 0.95 and specificity of 0.92 at the fasting time point and AUC of 0.99 (95% CI, 0.97–1.00), sensitivity of 0.95 and specificity of 0.95 at the 2-hour post-load time point.
Discussion
Prediabetes is a heterogeneous disorder of glucose metabolism characterized by IFG and/or IGT based on a 2-hour glucose load. Postpartum follow-up based on OGTT is complicated and depends on blood glucose levels at fasting and 2-hour post-load time points for the diagnosis of prediabetes.
In the current study, we conducted a follow-up of women with a history of GDM between six-week to six-month postpartum based on OGTT. Forty subjects had normal postpartum glucose tolerance, and 42 subjects were diagnosed with postpartum prediabetes, including seven with IFG, 31 with IGT, and four with both IFG and IGT. Plasma metabolite levels were significantly different between the prediabetes and control groups at the fasting and 2-hour post-load time points, respectively. The consistent detection of 15 differential metabolites in the prediabetes group at both the fasting and 2-hour post-load time points highlights their potential utility in assessing disease conditions and identifying biomarkers for diagnosis and postpartum follow-up screening. The metabolomic pathway analysis provides insights into potential pathophysiological mechanisms and etiological factors underlying prediabetes at early postpartum.
Consistent with several international studies[16-17], the current study found that the subjects were older in the prediabetes group than in the control group. Advanced maternal age is a known predictor for abnormal glucose tolerance between six to 12 week postpartum[18-19]. We also observed that the subjects in the prediabetes group showed higher 2-hour glucose values and the AUC of glucose during the pregnancy OGTT, compared with those in the control group, consistent with the findings from a systematic review by Moore et al[20], in which a higher 2-hour glucose level during pregnancy increased the risk of developing diabetes fourfold.
Previous studies on predicting the development of prediabetes and/or T2D have analyzed clinical markers and maternal serum or plasma metabolome during pregnancy or collected postpartum[9-11,21]. From a clinical perspective, there is a need to use only clinical markers that are easy to obtain during pregnancy to predict which women with a history of GDM are more likely to develop prediabetes postpartum. For example, Liu et al[10] found that the addition of metabolites at the gestation of approximately 28 weeks showed little improvement in predicting a postpartum disorder of glucose metabolism, compared with clinical factors alone. The AUC values of the clinical variables reported in previous studies[9-11] were also similar (ranging from 0.68 to 0.745) to those observed in the current study (0.71). In contrast, another study[21] with a small sample size and a longer follow-up period of nine years reported a slightly lower AUC for clinical markers (0.65). These results are relatively similar, notwithstanding the variations in the study populations, methodologies, and the clinical markers included. Therefore, clinical markers during pregnancy can be employed to preliminarily screen for individuals at high-risk of prediabetes.
The combination of metabolites and clinical markers is supposed to achieve a higher discriminating power than either used alone. Wang et al[11] identified the lipid species at 24 to 72 hours after delivery modestly improved the predictive performance for incident T2D (AUCmax ≤ 0.83) above classical risk factors across the first 15 years of follow-up. An American team, using the Study of Women, Infant Feeding, and Type 2 Diabetes after GDM Pregnancy (SWIFT) cohort and employing metabolomics and lipidomics, has established a series of predictive models for the progression from GDM to T2D, such as a panel of 12 lipids (AUC, 0.84)[22], a panel of 11 lipids (AUC, 0.739)[23], a signature of 20 metabolites (AUC, 0.88)[9] and a model with 10 analytes (AUC, 0.78)[24]. Given the cross-sectional study, we found that a panel of 15 metabolites showed a robust discriminative power for distinguishing between postpartum prediabetes and normal glucose tolerance at both the fasting and 2-hour post-load time points. However, future studies are warranted to validate the predictive power.
The metabolic profiling here revealed that energy metabolism and BCAAs played a role in the development of prediabetes in previous GDM women, which may gain insights into the underlying pathophysiology of the transition from GDM to prediabetes at the early postpartum period through the metabolic approach.
Both glycolysis and TCA cycle disruption characterize diabetes in multiple tissues[25]. The impaired insulin signaling may trigger inflammation via pyruvate accumulation[26-27]. The impaired TCA flux in insulin-resistant human skeletal muscle is one of the characteristics of the diabetic phenotype[28]. Patients with normal blood glucose but insulin resistance exhibit a 32% reduction in ATP production, compared with the control group[29]. At the 2-hour post-load time point, we observed an increase in pyruvate levels in prediabetes than in the control, while the TCA product fumarate decreased without changes in other intermediates. These imply that the excessive post-load glycolysis and possibly impaired TCA under insulin resistance may disrupt energy metabolism and elevate glucose.
Furthermore, plasma BCAA levels have been found to be correlated with insulin resistance[15,30], GDM[31], prediabetes[15], and T2D[9,15]. We previously also identified elevated BCAAs before and after delivery in GDM women using 1H-NMR[32]. The SWIFT cohort study observed that the elevated BCAA levels at six to nine weeks postpartum in women with GDM were strongly associated with an increased risk of developing T2D in the future[9]. In addition, Andersson-Hall et al[33] found higher fasting leucine levels in women with IGT than that with normal glucose tolerance at six years postpartum after GDM. In partial agreement with these findings, here we also found that leucine levels significantly increased at the 2-hour post-load time point in the prediabetes group than in the control group. Leucine is a potent activator of mTORC1, and persistent activation of mTORC1 may lead to or exacerbate insulin resistance[34-35].
The main strength of the current study lies in the identification of a panel of 15 metabolites that may distinguish between prediabetes and normal glucose tolerance at the early postpartum period in women with a history of GDM and could be considered for clinical application once validated in future studies.
The current study has some limitations. Our results may not be robust, because of a small sample size (n = 82). The cross-sectional approach forbids establishing causal links between the development of prediabetes and metabolic variations. Future larger cohorts will help to validate these findings with long-term effects.
In conclusion, the current study highlighted the metabolic changes associated with the transition from GDM to prediabetes in the early postpartum period, with a focus on energy metabolism and branched-chain amino acids. The identification of a panel of 15 metabolites with strong discriminative power for postpartum prediabetes suggests the potential for developing a metabolic profiling test for early risk assessment. While our findings require validation in larger, longitudinal studies, they provide a foundation for future research aimed at improving postpartum screening and early intervention strategies for women with a history of GDM. Such metabolic profiling could complement or potentially replace the current OGTT-based follow-up, addressing the challenges of low compliance and offering more precise risk stratification.
Acknowledgments
The authors acknowledge and appreciate all the volunteers who were involved in the study and our colleagues for their valuable efforts and comments on this paper.
Fundings
This work was supported by the National Natural Science Foundation of China (Grant No. 81800698), the Jiangsu Provincial Medical Key Discipline Cultivation Unit (Grant No. JSDW202241), the Research Project of Jiangsu Commission of Health (Grant No. H2023053), and Zhenjiang Science and the Technology Planning Project (Grant Nos. SH2023006 and SH2023008).
Siegel RL, Miller KD, Fuchs HE, et al. Cancer statistics, 2022[J]. CA Cancer J Clin, 2022, 72(1): 7–33. doi: 10.3322/caac.21708
[2]
Lin R, Yang S, Jia Y, et al. Epigenetic regulation of papillary thyroid carcinoma by long non-coding RNAs[J]. Semin Cancer Biol, 2022, 83: 253–260. doi: 10.1016/j.semcancer.2021.03.027
[3]
Lee TY, Lee S, Bae JS, et al. Distinct clinical manifestations of thyroid cancer after hematopoietic stem cell transplantation[J]. Ann Surg Oncol, 2019, 26(11): 3586–3592. doi: 10.1245/s10434-019-07442-4
[4]
La Vecchia C, Malvezzi M, Bosetti C, et al. Thyroid cancer mortality and incidence: a global overview[J]. Int J Cancer, 2015, 136(9): 2187–2195. doi: 10.1002/ijc.29251
[5]
Albero A, Lopéz JE, Torres A, et al. Effectiveness of chemotherapy in advanced differentiated thyroid cancer: a systematic review[J]. Endocr Relat Cancer, 2016, 23(2): R71–R84. doi: 10.1530/ERC-15-0194
[6]
Yu J, Deng Y, Liu T, et al. Lymph node metastasis prediction of papillary thyroid carcinoma based on transfer learning radiomics[J]. Nat Commun, 2020, 11(1): 4807. doi: 10.1038/s41467-020-18497-3
[7]
Jeppesen DK, Fenix AM, Franklin JL, et al. Reassessment of exosome composition[J]. Cell, 2019, 177(2): 428–445. e18.
[8]
Marar C, Starich B, Wirtz D. Extracellular vesicles in immunomodulation and tumor progression[J]. Nat Immunol, 2021, 22(5): 560–570. doi: 10.1038/s41590-021-00899-0
[9]
Luan W, Ding Y, Xi H, et al. Exosomal miR-106b-5p derived from melanoma cell promotes primary melanocytes epithelial-mesenchymal transition through targeting EphA4[J]. J Exp Clin Cancer Res, 2021, 40(1): 107. doi: 10.1186/s13046-021-01906-w
[10]
Thakur BK, Zhang H, Becker A, et al. Double-stranded DNA in exosomes: a novel biomarker in cancer detection[J]. Cell Res, 2014, 24(6): 766–769. doi: 10.1038/cr.2014.44
[11]
Wang Y, Liu J, Ma J, et al. Exosomal circRNAs: biogenesis, effect and application in human diseases[J]. Mol Cancer, 2019, 18(1): 116. doi: 10.1186/s12943-019-1041-z
[12]
Dai J, Su Y, Zhong S, et al. Exosomes: key players in cancer and potential therapeutic strategy[J]. Signal Transduct Target Ther, 2020, 5(1): 145. doi: 10.1038/s41392-020-00261-0
[13]
Li X, Yang L, Chen L. The biogenesis, functions, and challenges of circular RNAs[J]. Mol Cell, 2018, 71(3): 428–442. doi: 10.1016/j.molcel.2018.06.034
[14]
Liu C, Chen L. Circular RNAs: characterization, cellular roles, and applications[J]. Cell, 2022, 185(12): 2016–2034. doi: 10.1016/j.cell.2022.04.021
[15]
Xu X, Jing J. Advances on circRNAs contribute to carcinogenesis and progression in papillary thyroid carcinoma[J]. Front Endocrinol (Lausanne), 2021, 11: 555243. doi: 10.3389/fendo.2020.555243
[16]
Xu B, Huang X, Yan Y, et al. Analysis of expression profiles and bioinformatics suggests that plasma exosomal circular RNAs may be involved in ischemic stroke in the Chinese Han population[J]. Metab Brain Dis, 2022, 37(3): 665–676. doi: 10.1007/s11011-021-00894-2
[17]
Xu M, Lin B, Zheng D, et al. LEM domain containing 1 promotes thyroid cancer cell proliferation and migration by activating the Wnt/β-catenin signaling pathway and epithelial-mesenchymal transition[J]. Oncol Lett, 2021, 21(6): 442. doi: 10.3892/ol.2021.12703
[18]
Sa R, Liang R, Qiu X, et al. IGF2BP2-dependent activation of ERBB2 signaling contributes to acquired resistance to tyrosine kinase inhibitor in differentiation therapy of radioiodine-refractory papillary thyroid cancer[J]. Cancer Lett, 2022, 527: 10–23. doi: 10.1016/j.canlet.2021.12.005
[19]
Rusciano MR, Salzano M, Monaco S, et al. The Ca2+-calmodulin-dependent kinase II is activated in papillary thyroid carcinoma (PTC) and mediates cell proliferation stimulated by RET/PTC[J]. Endocr Relat Cancer, 2010, 17(1): 113–123. doi: 10.1677/ERC-09-0214
Zhao Q, Ming J, Liu C, et al. Multifocality and total tumor diameter predict central neck lymph node metastases in papillary thyroid microcarcinoma[J]. Ann Surg Oncol, 2013, 20(3): 746–752. doi: 10.1245/s10434-012-2654-2
[22]
Caliskan O, Unlu MT, Yanar C, et al. Predictive factors affecting the development of lateral lymph node metastasis in papillary thyroid cancer[J]. Sisli Etfal Hastan Tip Bul, 2023, 57(3): 312–319. https://pubmed.ncbi.nlm.nih.gov/37900340/
[23]
Haugen BR, Alexander EK, Bible KC, et al. 2015 American thyroid association management guidelines for adult patients with thyroid nodules and differentiated thyroid cancer: the American thyroid association guidelines task force on thyroid nodules and differentiated thyroid cancer[J]. Thyroid, 2016, 26(1): 1–133. doi: 10.1089/thy.2015.0020
[24]
Dai L, Hu W, Jiang H, et al. The diagnostic potential of two exosome-derived circRNAs for papillary thyroid cancer[J]. Int J Clin Oncol, 2023, 28(11): 1461–1474. doi: 10.1007/s10147-023-02400-3
[25]
Lin Q, Qi Q, Hou S, et al. Exosomal circular RNA hsa_circ_007293 promotes proliferation, migration, invasion, and epithelial-mesenchymal transition of papillary thyroid carcinoma cells through regulation of the microRNA-653-5p/paired box 6 axis[J]. Bioengineered, 2021, 12(2): 10136–10149. doi: 10.1080/21655979.2021.2000745
[26]
Paskeh MDA, Entezari M, Mirzaei S, et al. Emerging role of exosomes in cancer progression and tumor microenvironment remodeling[J]. J Hematol Oncol, 2022, 15(1): 83. doi: 10.1186/s13045-022-01305-4
[27]
De Falco V, Castellone MD, De Vita G, et al. RET/papillary thyroid carcinoma oncogenic signaling through the Rap1 small GTPase[J]. Cancer Res, 2007, 67(1): 381–390. doi: 10.1158/0008-5472.CAN-06-0981
[28]
Lu Z, Wen D, Wei W, et al. Silencing of PPM1D inhibits cell proliferation and invasion through the p38 MAPK and p53 signaling pathway in papillary thyroid carcinoma[J]. Oncol Rep, 2020, 43(3): 783–794. doi: 10.3892/or.2020.7458
[29]
Jiang H, Chen X, Sun H, et al. Tumor promoting effects of glucagon receptor: a promising biomarker of papillary thyroid carcinoma via regulating EMT and P38/ERK pathways[J]. Hum Cell, 2020, 33(1): 175–184. doi: 10.1007/s13577-019-00284-y
[30]
Kim YR, Byun HS, Won M, et al. Modulatory role of phospholipase D in the activation of signal transducer and activator of transcription (STAT)-3 by thyroid oncogenic kinase RET/PTC[J]. BMC Cancer, 2008, 8: 144. doi: 10.1186/1471-2407-8-144
[31]
Song H, Qiu Z, Wang Y, et al. HIF-1α/YAP signaling rewrites glucose/iodine metabolism program to promote papillary thyroid cancer progression[J]. Int J Biol Sci, 2023, 19(1): 225–241. doi: 10.7150/ijbs.75459
[32]
Zhan S, Wang T, Li J, et al. Asporin interacts with HER2 to promote thyroid cancer metastasis via the MAPK/EMT signaling pathway[J]. Front Oncol, 2022, 12: 762180. doi: 10.3389/fonc.2022.762180
[33]
Cao Z, Zhang Z, Tang X, et al. Comprehensive analysis of tissue proteomics in patients with papillary thyroid microcarcinoma uncovers the underlying mechanism of lymph node metastasis and its significant sex disparities[J]. Front Oncol, 2022, 12: 887977. doi: 10.3389/fonc.2022.887977
[34]
Mirshahidi S, Yuan IJ, Simental A, et al. Targeting tumor microenvironment akt signaling represents a potential therapeutic strategy for aggressive thyroid cancer[J]. Int J Mol Sci, 2023, 24(6): 5471. doi: 10.3390/ijms24065471
[35]
Chen L. The expanding regulatory mechanisms and cellular functions of circular RNAs[J]. Nat Rev Mol Cell Biol, 2020, 21(8): 475–490. doi: 10.1038/s41580-020-0243-y
[36]
Hansen TB, Jensen TI, Clausen BH, et al. Natural RNA circles function as efficient microRNA sponges[J]. Nature, 2013, 495(7441): 384–388. doi: 10.1038/nature11993
[37]
Yao Y, Chen X, Yang H, et al. Hsa_circ_0058124 promotes papillary thyroid cancer tumorigenesis and invasiveness through the NOTCH3/GATAD2A axis[J]. J Exp Clin Cancer Res, 2019, 38(1): 318. doi: 10.1186/s13046-019-1321-x
[38]
Zhang D, Tao L, Xu N, et al. CircRNA circTIAM1 promotes papillary thyroid cancer progression through the miR-646/HNRNPA1 signaling pathway[J]. Cell Death Discov, 2022, 8(1): 21. doi: 10.1038/s41420-021-00798-1
[39]
Manzella L, Stella S, Pennisi MS, et al. New insights in thyroid cancer and p53 family proteins[J]. Int J Mol Sci, 2017, 18(6): 1325. doi: 10.3390/ijms18061325
[40]
Scheffel RS, Dora JM, Maia AL. BRAF mutations in thyroid cancer[J]. Curr Opin Oncol, 2022, 34(1): 9–18. doi: 10.1097/CCO.0000000000000797
[41]
Leandro-García LJ, Landa I. Mechanistic insights of thyroid cancer progression[J]. Endocrinology, 2023, 164(9): bqad118. doi: 10.1210/endocr/bqad118
[42]
Zhang Z, Liu Z, Ren W, et al. The miR-200 family regulates the epithelial-mesenchymal transition induced by EGF/EGFR in anaplastic thyroid cancer cells[J]. Int J Mol Med, 2012, 30(4): 856–862. doi: 10.3892/ijmm.2012.1059
[43]
Hua K, Jin J, Zhang H, et al. MicroRNA-7 inhibits proliferation, migration and invasion of thyroid papillary cancer cells via targeting CKS2[J]. Int J Oncol, 2016, 49(4): 1531–1540. doi: 10.3892/ijo.2016.3660
[44]
Zhang H, Duan J, Qu Y, et al. Onco-miR-24 regulates cell growth and apoptosis by targeting BCL2L11 in gastric cancer[J]. Protein Cell, 2016, 7(2): 141–151. doi: 10.1007/s13238-015-0234-5
[45]
Rose MM, Espinoza VL, Hoff KJ, et al. BCL2L11 induction mediates sensitivity to src and MEK1/2 inhibition in thyroid cancer[J]. Cancers (Basel), 2023, 15(2): 378. doi: 10.3390/cancers15020378
[46]
Dou X, Xia F, Li X. Circ_0003747 promotes thyroid cancer progression by sponging miR-338-3p to upregulate PLCD3 expression[J]. Epigenetics, 2023, 18(1): 2210339. doi: 10.1080/15592294.2023.2210339
[47]
Lin L, Wen J, Lin B, et al. Phospholipase C Delta 3 inhibits apoptosis and promotes proliferation, migration, and invasion of thyroid cancer cells via Hippo pathway[J]. Acta Biochim Biophys Sin (Shanghai), 2021, 53(4): 481–491. doi: 10.1093/abbs/gmab016
[48]
Han Y, Yu X, Yin Y, et al. Identification of potential BRAF inhibitor joint therapy targets in PTC based on WGCAN and DCGA[J]. J Cancer, 2021, 12(6): 1779–1791. doi: 10.7150/jca.51551
[49]
Lee JH, Horak CE, Khanna C, et al. Alterations in Gemin5 expression contribute to alternative mRNA splicing patterns and tumor cell motility[J]. Cancer Res, 2008, 68(3): 639–644. doi: 10.1158/0008-5472.CAN-07-2632
[50]
Salib A, Jayatilleke N, Seneviratne JA, et al. MYCN and SNRPD3 cooperate to maintain a balance of alternative splicing events that drives neuroblastoma progression[J]. Oncogene, 2024, 43(5): 363–377. doi: 10.1038/s41388-023-02897-y
Table
1.
Postpartum demographic characteristics of the subjects
Characteristics
Control (n=40)
Prediabetes (n=42)
P-value
Age (years, median [Q1, Q3])
31 (29–33)
33 (30–35)
0.010*
Multipara (n [%])
8 (20.0%)
16 (38.1%)
0.072
Family history of diabetes (n [%])
17 (42.5%)
17 (40.5%)
0.852
Exclusive breastfeeding (n [%])
20 (50%)
16 (38.1%)
0.278
Weight (kg, mean ± SD)
60.4±9.0
61.4±7.2
0.594
PWL (kg, mean ± SD)
8.3±3.0
8.6±2.8
0.572
PWR (kg, mean ± SD)
3.5±5.0
3.4±3.8
0.984
BMI (kg/m2, mean ± SD)
23.0±3.1
23.4±2.5
0.612
SBP (mmHg, mean ± SD)
112±10
112±12
0.959
DBP (mmHg, mean ± SD)
70±7
71±9
0.546
Waist circumference (cm, mean ± SD)
82.1±8.4
82.9±9.2
0.698
Hip circumference (cm, mean ± SD)
94.7±5.9
96.0±6.1
0.308
Waist-hip ratio (mean ± SD)
0.87±0.05
0.86±0.06
0.718
Data are expressed as means ± standard deviation (SD), medians (interquartile ranges), or numbers (percentages). Differences between the two groups were tested by the Chi-square test for categorical variables, the Student's t-test for normally distributed continuous variables, and the Mann-Whitney test for non-normally distributed variables. *P < 0.05. Abbreviations: BMI, body mass index; SBP, systolic blood pressure; DBP, diastolic blood pressure.
Table
2.
Metabolites detected between the groups at different time points
Metabolites
Fasting
Metabolites
Two-hour post-load
log2(FC)
P-value
log2(FC)
P-value
(E)-cinnamaldehyde
0.64
**
(R)-carnitine
0.36
**
(S)-allantoin
0.15
*
1,3,7-Trimethyluric acid
−0.22
*
(S)-lactatea
−0.35
***
11-deoxyprostaglandin E1
−0.55
*
1,3,7-Trimethyluric acid
−0.24
**
2(5H)-Furanone
0.21
***
10-Hydroxydecanoic acid
0.25
*
2-Butynedioic acid
0.03
*
11-deoxyprostaglandin E1
−0.35
**
2-deoxy-D-glucosea
−0.31
***
2-aminoacrylic acid
−0.17
**
2-oxoadipic acid
0.15
***
2-deoxy-D-glucosea
−0.16
**
5,6-dihydroxyindole
−0.35
***
3,4-dihydroxymandelic acid
0.09
*
5-acetamido-6-formamido-3-methyluracil
0.49
***
4-hydroxyphenylacetic acid
0.12
*
7-methylxanthine
0.23
*
7,9-dihydro-1H-purine-2,6,8(3H)-trione
−0.28
**
acetoin
0.11
*
7-methylxanthine
−0.25
*
acrylate
0.21
***
allantoin
−0.43
**
allantoin
−0.4
*
alpha-Ketocaproic acid
0.21
*
alpha-D-galactose
0.22
***
arachidonic acid
0.44
***
benzene-1,2,4-triol
0.23
***
benzoic acid
0.17
**
butane-2,3-dione
0.25
***
beta-alanine
−0.45
**
chenodeoxycholic acidc
0.24
*
cholic acidc
−0.32
*
cholic acidc
−0.27
***
D-aspartic acid
−0.49
***
corticosterone
0.24
**
D-glucono-1,5-lactone
−0.48
***
cortisone
0.25
*
dodecanedioic acid
0.42
*
decanoate
0.38
*
dodecanoate
−0.56
*
decanoic acid
−0.28
***
D-Pyroglutamic acidb
0.19
**
D-fructofuranose 6-phosphate
0.14
**
D-tagatopyranose
−0.37
**
D-glucuronate
0.16
*
dUMP
−0.63
***
D-Pyroglutamic acidb
0.07
*
equol
0.34
*
dUMP
−0.25
*
estrone
0.6
*
fumaratea
−0.15
***
fumaratea
−0.07
*
galactose
0.51
***
fumaric acid
−0.24
**
glutaric acid
0.17
***
gamma-aminobutyric acidb
0.09
**
glycine betaine
0.1
*
homovanillic acid
−0.06
**
glycocholic acidbc
−0.2
*
hyocholic acidc
0.25
*
glycolate
0.28
***
hypoxanthine
−0.66
***
glycolic acid
−0.46
***
isocitric acida
0.23
**
glyoxylic acid
0.26
***
L-glutamic acidb
0.26
**
hyocholic acidc
0.72
*
linoleate
0.23
*
hyodeoxycholic acid
0.12
*
L-proline
−0.23
***
indol-3-ylacetaldehyde
0.61
***
L-Rhamnose
−0.11
*
indole-3-acetic acid
0.35
***
L-threonic acid
−0.44
***
L-arabinitol
0.13
***
L-tyrosine
−0.35
**
L-glutamineb
−0.08
*
maltose
−0.44
***
L-leucine
0.28
**
methylmalonic acid
−0.51
***
L-Rhamnose
−0.44
***
N-acetyl-L-alanine
0.22
*
L-sorbose
0.23
***
N-acetylneuraminic acid
−0.28
**
Methionine sulfoxide
0.13
*
N-benzoylglycine
−0.23
*
methylmalonate
0.17
***
N-cinnamoylglycine
−0.08
**
myo-inositol
0.3
***
octadecanoic acid
0.21
**
N-acetyl-L-proline
0.24
***
octanoic acid
0.17
*
N-acetylneuraminic acid
−0.38
*
oxomalonic acid
−0.15
*
N-cinnamoylglycine
−0.18
***
p-hydroxyhippuric acid
−0.41
***
octanoic acid
0.39
**
Purine
−0.49
*
oxomalonic acid
−0.19
***
pyridoxal
−0.51
*
pyruvatea
0.26
***
sucrosea
−0.35
***
quinolin-2-ol
−0.27
***
uric acid
−0.51
***
S-sulfo-L-cysteine
0.29
*
xylitol
−0.21
*
theobromine
0.32
***
thymidine
−0.51
***
tyramine sulfate
0.15
**
xylitol
−0.29
**
aMetabolites associated with energy metabolism. bMetabolites associated with glutamate metabolism. cMetabolites associated with bile acid metabolism.Concentration changes of metabolites are indicated by a color gradient, with red for increased and blue for decreased levels, based on the log-transformed fold change (log2[FC]). The significance was determined using P-values adjusted by the Benjamini-Hochberg method for either Student's t-test or Mann-Whitney test. *P < 0.05, **P < 0.01, and ***P < 0.001.
Authors and Reviewers
DownLoad:
