Gestational diabetes mellitus (GDM), defined as glucose intolerance with onset or first recognition during pregnancy [
Comprehensive metabolic profiling (metabolomics) has provided unique insights into the mechanisms of insulin resistance. For example, branched-chain amino acids (BCAAs) and the aromatic amino acids (AAAs) are strongly associated with metabolic disease [
Recently, we characterized the metabolic transition profile from pregnancy to postpartum by mass spectrometry-based metabolomics in women with GDM [
3-Hydroxy-isobutyrate (3-HIB), a catabolic intermediate of the BCAA valine, may be released from muscle cells and regulates transendothelial transport and muscle cell uptake of FFAs [
In this study, we compared the metabolic profiles of women who had impaired glucose tolerance (IGT) or T2DM or were normoglycaemic 6 years after GDM. We hypothesized that alterations in BCAA metabolism interact with lipid metabolism and glucose oxidative pathways to induce insulin resistance and diabetes.
The study was approved by the ethics committee at the University of Gothenburg (402-08/750-15). Informed consent was obtained from all participants. Data were collected retrospectively from women in the Gothenburg area who were diagnosed with GDM from 2005 to 2009 (
Flow chart of study population and pregnancy data. Age, age at partus; BMI, BMI at start of pregnancy (kg/m2); ethnic, % Nordic ethnicity; glucose, fasting p-glucose (mM) at time of GDM diagnosis; diagnose, gestational age at time of GDM diagnosis; and insulin req, required insulin treatment during GDM pregnancy.
At follow-up, fasting blood samples were collected for all women, and those who had not previously been diagnosed with diabetes or who did not have a diagnostic fasting plasma glucose value underwent a 2 h 75 g OGTT. Venous blood for analysis of plasma glucose and serum insulin was collected at 0, 30, 60, 90, and 120 minutes. Capillary glucose was measured at 0 and 120 minutes as a backup in case venous sampling was unsuccessful. Fasting serum and plasma blood samples were used for further analysis.
Anthropometric measurements (weight, height, waist, and hip circumference) and resting blood pressure were determined, and participants were asked to complete questionnaires during the visit. A self-administered dietary questionnaire was used to assess energy intake (EI) during the 3 previous months [
Based on fasting and 2 h glucose values from the OGTT at the 6-year follow-up, three groups were created: normal glucose tolerance (NGT), IGT (including impaired fasting glucose or impaired glucose tolerance), and T2DM (including subjects with previously diagnosed T2DM). Classification was based on venous glucose levels according to the 1999 WHO guidelines [
HbA1c was analysed immediately with a point-of-care analyser (Afinion AS100; Axis-Shield, Oslo, Norway). Glucose, insulin, cholesterol (total, low-density lipoprotein (LDL), high-density lipoprotein (HDL)), triglycerides, leptin, adiponectin, C-reactive protein, and FFAs were analysed at the accredited Clinical Chemistry Laboratory, Sahlgrenska University Hospital (International Standard ISO 15189:2007). ELISA was used to measure adiponectin (Human Adiponectin ELISA kit, Millipore, Billerica, MA; interassay coefficient of variation, 7.0% at 10.5 mg/L) and leptin (Human Leptin Quantikine, R&D Systems; interassay coefficient of variation, 8.0% at 9
Anthropometry, blood pressure (BP), clinical blood measurements, and self-reported lifestyle data at follow-up 6 years after GDM pregnancy.
Variable | NGT | IGT | T2DM | ||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Mean | SD | Mean | SD | Mean | SD | IGT versus NGT | T2D versus NGT | T2D versus IGT | IGT versus NGT | T2D versus NGT | T2D versus IGT | ||||
BMI (kg/m2) | 139 | 26.1 | 4.8 | 46 | 29.4 | 4.5 | 44 | 29.1 | 5.8 | <0.001 | 0.002 | ||||
BMI change (kg/m2) | 139 | −0.4 | 3.0 | 46 | 1.8 | 3.1 | 44 | −0.4 | 3.5 | <0.001 | 0.003 | 0.001 | |||
Waist (cm) | 137 | 87.5 | 10.8 | 45 | 94.2 | 16.2 | 44 | 94.6 | 14.5 | 0.007 | 0.004 | ||||
Hip (cm) | 137 | 102.3 | 9.7 | 45 | 108.4 | 9.1 | 44 | 107.1 | 11.8 | 0.002 | 0.02 | ||||
Waist : hip ratio | 137 | 0.9 | 0.1 | 45 | 0.9 | 0.1 | 44 | 0.9 | 0.1 | ||||||
Body fat (%) | 59 | 32.9 | 7.9 | 16 | 40.5 | 6.6 | 12 | 38.5 | 13.2 | 0.007 | |||||
Fat mass (kg) | 59 | 23.4 | 9.5 | 16 | 32.3 | 10.7 | 12 | 31.8 | 15.8 | 0.01 | 0.04 | ||||
Fat-free mass (kg) | 59 | 45.6 | 5.4 | 16 | 46.0 | 6.7 | 12 | 45.9 | 6.0 | ||||||
BP systolic (mmHg) | 138 | 113.9 | 12.8 | 46 | 122.9 | 14.0 | 44 | 119.2 | 14.2 | <0.001 | 0.02 | ||||
BP diastolic (mmHg) | 138 | 74.9 | 9.6 | 46 | 78.6 | 9.8 | 44 | 78.1 | 10.6 | ||||||
b-HbA1c (mmol/mol) | 139 | 37.1 | 3.7 | 45 | 38.4 | 3.8 | 44 | 52.6 | 18.8 | <0.001 | <0.001 | <0.001 | <0.001 | ||
p-glucose fasting (mM) | 138 | 5.4 | 0.4 | 43 | 6.1 | 0.4 | 44 | 8.2 | 3.5 | 0.02 | <0.001 | <0.001 | 0.02 | <0.001 | <0.001 |
p-glucose 2 h (mM) | 128 | 5.5 | 1.1 | 42 | 7.8 | 1.6 | 12 | 11.8 | 3.9 | <0.001 | <0.001 | <0.001 | <0.001 | <0.001 | <0.001 |
s-insulin fasting (mU/L) | 134 | 8.0 | 4.2 | 43 | 14.4 | 9.5 | 44 | 12.8 | 7.4 | <0.001 | <0.001 | <0.001 | 0.003 | ||
s-insulin 2 h (mU/L) | 121 | 44.7 | 37.7 | 40 | 87.1 | 57.1 | 12 | 85.3 | 58.4 | <0.001 | 0.008 | <0.001 | 0.01 | ||
HOMA-IR | 133 | 1.9 | 1.0 | 43 | 3.8 | 2.7 | 43 | 4.9 | 3.4 | <0.001 | <0.001 | 0.04 | <0.001 | <0.001 | 0.004 |
s-total cholesterol (mM) | 137 | 4.7 | 0.8 | 45 | 4.7 | 0.8 | 43 | 4.7 | 0.7 | ||||||
s-HDL (mM) | 138 | 1.5 | 0.3 | 45 | 1.3 | 0.3 | 44 | 1.4 | 0.4 | 0.001 | 0.04 | 0.04 | |||
s-LDL (mM) | 138 | 2.9 | 0.8 | 45 | 3.1 | 0.7 | 44 | 3.0 | 0.6 | ||||||
s-triglycerides (mM) | 137 | 0.9 | 0.5 | 45 | 1.2 | 0.5 | 44 | 1.4 | 1.4 | <0.001 | 0.002 | ||||
s-leptin (mg/L) | 134 | 18.2 | 11.8 | 42 | 29.3 | 12.9 | 44 | 24.4 | 22.4 | <0.001 | 0.04 | 0.03 | |||
s-adiponectin (ug/L) | 137 | 17.8 | 14.5 | 41 | 18.1 | 15.9 | 44 | 14.2 | 12.9 | ||||||
Leptin : adiponectin 10−3 | 137 | 1.86 | 1.95 | 41 | 2.83 | 2.52 | 44 | 3.51 | 3.67 | 0.001 | 0.02 | ||||
s-C-reactive protein (mg/L) | 137 | 2.1 | 3.6 | 43 | 3.1 | 3.1 | 44 | 5.5 | 8.2 | <0.001 | 0.008 | 0.007 | |||
p-free fatty acids (mM) | 138 | 0.43 | 0.18 | 41 | 0.56 | 0.20 | 44 | 0.59 | 0.22 | 0.001 | <0.001 | 0.001 | <0.001 | ||
ADIPO-IR (mM·pM) | 134 | 24.1 | 17.6 | 40 | 53.6 | 38.2 | 44 | 55.7 | 40.5 | <0.001 | <0.001 | <0.001 | <0.001 | ||
Physical activity | 120 | 4.0 | 1.1 | 38 | 3.3 | 1.0 | 34 | 3.4 | 1.3 | 0.008 | 0.02 | 0.009 | 0.02 | ||
Energy intake (kcal) | 88 | 2390 | 561 | 31 | 2366 | 735 | 17 | 2258 | 610 | ||||||
Protein intake (g/kg) | 88 | 1.4 | 0.4 | 31 | 1.2 | 0.4 | 17 | 1.4 | 0.5 | ||||||
Fat intake (g/kg) | 88 | 1.6 | 0.5 | 31 | 1.3 | 0.4 | 17 | 1.5 | 0.5 | ||||||
Carbohydrate intake (g/kg) | 88 | 3.7 | 1.2 | 31 | 3.2 | 1.3 | 17 | 3.4 | 1.3 |
The women were divided into glucose tolerance groups based on follow-up OGTT. Statistical analyses were done with ANOVA (
224 fasting serum samples were available for NMR analysis after all other blood analysis was performed. NMR data were acquired on a Bruker 800 MHz spectrometer equipped with a 3 mm TCI CryoProbe and a cooled SampleJet sample changer using the “zgespe” pulse sequence encoding a one-dimensional 1H experiment with suppression of water and J-modulation by excitation sculpting and perfect echo, respectively, and T2 filtering with a CPMG pulse train. For each sample, 128 scans were collected into 65k points with an acquisition time of 2.04 seconds, relaxation delay of 3 seconds, sweep width of 20 ppm, and a total CPMG pulse train of 193 ms (total experimental time, 12 minutes 4 seconds). The temperature was kept at 6°C in the sample changer and at 25°C during acquisition. Spectral data were processed with TopSpin 3.5pl6 (Bruker BioSpin), zero filling once, and 0.3 Hz line broadening before Fourier transform. Representative samples for the different groups were subjected to two-dimensional NMR data acquisition using 1H-1H TOCSY (“mlevgpphw5”), 1H,13C-HSQC (“hsqcetgpsisp2.2”), and two-dimensional Jres (“jresgpprqf”) experiments. Two-dimensional results for 3-HIB are shown in Supplementary Materials (available
Pregnancy data were analysed with one-way ANOVA and post hoc Tukey analysis or chi-square test (binary variables). Follow-up measurements were analysed with ANOVA and ANCOVA (BMI as covariate). To predict the likelihood of T2DM from pregnancy data, variables were analysed by binary regression (outcome T2DM, yes/no). For association of follow-up variables with insulin resistance, a stepwise forward linear regression model was built with HOMA-IR as the dependant variable. All variables in Table
NMR metabolites at follow-up 6 years after GDM pregnancy.
% | NGT | IGT | T2DM | |||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
Mean | SD | Mean | SD | Mean | SD | IGT versus NGT | T2D versus NGT | T2D versus IGT | IGT versus NGT | T2D versus NGT | T2D versus IGT | |
Phenylalanine | 100.0 | 45.4 | 121.8 | 43.8 | 117.1 | 49.1 | 0.02 | |||||
Tyrosine | 100.0 | 25.1 | 109.6 | 26.9 | 111.6 | 26.7 | 0.03 | |||||
Glucose | 100.0 | 20.7 | 112.1 | 18.1 | 159.5 | 79.6 | <0.001 | <0.001 | <0.001 | <0.001 | ||
Mannose | 100.0 | 25.9 | 113.4 | 26.2 | 144.3 | 59.5 | <0.001 | <0.001 | <0.001 | <0.001 | ||
Glycerol | 100.0 | 43.1 | 127.5 | 48.6 | 140.7 | 71.1 | 0.006 | <0.001 | 0.007 | <0.001 | ||
Glycine | 100.0 | 27.0 | 88.1 | 19.1 | 89.6 | 25.6 | 0.02 | 0.05 | 0.04 | |||
Citrate | 100.0 | 31.0 | 111.5 | 23.9 | 117.3 | 25.7 | 0.002 | 0.002 | ||||
Pyruvate | 100.0 | 80.6 | 141.3 | 78.2 | 119.4 | 73.7 | 0.008 | 0.01 | ||||
Acetoacetate | 100.0 | 105.8 | 115.7 | 103.9 | 253.1 | 348.8 | <0.001 | 0.001 | <0.001 | 0.001 | ||
3-HIB | 100.0 | 54.7 | 128.2 | 52.7 | 131.5 | 75.5 | 0.02 | 0.006 | 0.005 | 0.002 | ||
Valine | 100.0 | 19.2 | 108.2 | 15.1 | 109.5 | 21.6 | 0.04 | 0.01 | 0.05 | |||
Isoleucine | 100.0 | 22.1 | 109.0 | 19.0 | 115.7 | 28.5 | <0.001 | 0.001 | ||||
Leucine | 100.0 | 18.2 | 111.2 | 13.6 | 113.4 | 22.7 | 0.002 | <0.001 | 0.01 | 0.001 |
Metabolites significantly different between groups are shown. Data is expressed as % of NGT. Statistics are performed using ANOVA (
After the follow-up OGTT, 58% had NGT (
By logistic regression, the strongest prediction model for development of T2DM included fasting glucose at GDM diagnosis (
Anthropometric and clinical data at the 6-year follow-up are shown in Table
As expected, glucose, insulin, and HOMA-IR were higher in the IGT and T2DM groups even after BMI adjustment. HbA1c was significantly increased in T2DM only. HDL was lower and leptin was higher in the IGT and T2DM groups than in the NGT group, but only the differences between IGT and NGT were significant after BMI adjustment. FFA and ADIPO-IR, reflecting inability of insulin to suppress peripheral lipolysis, were higher in the IGT and T2DM groups, and the difference was highly significant even after BMI adjustment; there were no differences in total cholesterol or LDL. The leptin : adiponectin ratio and C-reactive protein (CRP) were both higher in T2DM, independent of BMI. Physical activity was higher in the NGT group than in the other groups, even after BMI adjustment, and correlated with several metabolic parameters (BMI, waist circumference, hip circumference, HbA1c, fasting insulin, 2-hour OGTT glucose, HOMA-IR, HDL, leptin, ADIPO-IR, and CRP). After adjustment for BMI, significant correlations remained for hip circumference (
NMR metabolomics of serum samples from the 6-year follow-up showed significant differences between groups (Table
Map of metabolites that differed significantly between groups after BMI adjustment 6 years after a GDM pregnancy. Values are percent of NGT (mean ± SEM).
Figure
Correlations between glucose (a) and lipid (b) metabolism measurements and BCAAs, 3-HIB, and AAAs 6 years after GDM pregnancy. All three groups (NGT, IGT, and T2DM) were included. Only correlations that were significant after BMI adjustment are shown (
For AAAs, Tyr correlated with fasting insulin (
Other group differences included for T2DM increased mannose, acetoacetate, glycerol, and citrate and for IGT increased pyruvate and glycerol and decreased glycine.
To predict HOMA-IR variance, follow-up variables that correlated with HOMA-IR but are not direct measurements of glucose metabolism were analysed in a regression model. These included BMI, BMI change, age, blood pressure, waist and hip circumference, waist : hip ratio, triglycerides, cholesterol, HDL, LDL, CRP, leptin, adiponectin, leptin : adiponectin ratio, FFA, phenylalanine, tyrosine, glycine, valine, leucine, isoleucine, 3-HIB, and physical activity. After stepwise analysis, the model contained eight variables that gave an adjusted
Stepwise linear regression model with HOMA-IR as dependent variable and follow-up variables as independent variables.
Variables included in model | ||
---|---|---|
s-triglycerides (mM) | 0.35 | <0.001 |
Waist (cm) | 0.31 | <0.001 |
3-HIB (%) | 0.19 | <0.001 |
Age (y) | −0.18 | <0.001 |
s-C-reactive protein (mg/L) | 0.15 | 0.008 |
Diastolic BP (mmHg) | 0.13 | 0.01 |
Physical activity | −0.12 | 0.01 |
BMI change (kg/m2) | 0.11 | 0.02 |
Out of 22 variables, 8 remained in the final model to predict HOMA-IR variance;
Eight women were diagnosed with T1DM 6 years after GDM. This group had higher levels of acetoacetate and acetate than the NGT group (Figure
3-HIB and ketone concentrations in T1DM (
This study shows that the metabolic profile in women with IGT or T2DM differs substantially from that of women with NGT 6 years after GDM. At follow-up, 22% of the women had developed diabetes (19% T2DM and 3% T1DM) and 19% had IGT. Women with IGT and diabetes had an increased clinical risk profile (higher fasting glucose and insulin treatment during pregnancy) and developed diabetes at an earlier gestational age than those who remained normoglycaemic after pregnancy. These clinical risk factors, which have been shown in other follow-up studies [
The IGT and T2DM groups had significant differences in amino acids. AAAs and BCAAs were increased and glycine was decreased. The increases for T2DM in leucine, isoleucine, valine, and 3-HIB were significant after BMI adjustment, but those in tyrosine and phenylalanine were not. The BMI-independent differences in precursors and intermediates such as mannose, glucose, glycine, pyruvate, and citrate point to a possible dysregulation of glycolysis and the TCA-cycle. Changes in triglycerides, glycerol, and FFA suggest dysfunctional fat metabolism. In regression models, the risk factors mainly associated with HOMA-IR are in order of importance triglycerides, waist circumference, 3-HIB, age, CRP, diastolic blood pressure, physical activity, and change in BMI from pregnancy to follow-up.
Disturbed AA metabolism has long been considered a feature of obesity and associated metabolic disease; obese people have higher blood levels of the BCAAs (leucine, isoleucine, and valine) and AAA (phenylalanine and tyrosine) but lower levels of glycine than lean people [
BCAAs differ from other AAs in that they are initially catabolized primarily in skeletal muscle not in liver [
One important step in valine metabolism is generation of 3-HIB, which can be released from tissues. Elevated plasma concentrations of 3-HIB in 3-day fasted subjects and in T1DM patients fasted overnight reflect augmented valine degradation from net body protein breakdown [
Increased BCAA and 3-HIB levels can either be explained by increased protein breakdown and/or decreased catabolism of BCAAs and 3-HIB. In a recent study, several genes involved in BCAA catabolism were downregulated in muscle from insulin-resistant subjects [
American Diabetes Association guidelines for postpartum glucose testing were revised in 2017 to include a 75 g OGTT at 4–12 weeks postpartum and thereafter every 1–3 years for women with a prior GDM diagnosis, and more frequent testing if screening results fall within the prediabetes ranges [
Pharmaceutical intervention and lifestyle modifications are as effective in delaying or preventing the onset of T2DM after a GDM pregnancy as they are in other cases of reduced glucose tolerance [
With the availability and advancement of methods such as NMR and mass spectrometry, several other nonglucose metabolites associated with insulin resistance have been identified, such as 2-hydroxybutyrate (interestingly also linked to amino acid metabolism and the TCA cycle in similarity with 3-HIB), lipid signalling molecules, and fatty acids [
A limitation of the study is a potential recruitment bias out of the total Gothenburg GDM population, possibly due to language difficulties in the non-Nordic ethnic population. However, this factor is not likely to affect comparisons between glucose tolerance groups or correlations of metabolic parameters.
Metabolic risk scores that can identify women at the highest risk for transitioning from GDM to IGT and T2DM are needed. A cost-effective and easy-to-use strategy must be adopted by healthcare givers to reach these women and offer individualized prevention programs. A single blood sample taken annually to analyse one or a few metabolites, in addition to well-known clinical risk factors, would be beneficial but is currently unavailable. Thus, 3-HIB is an interesting metabolite warranting further evaluation as one of a suite of markers for insulin resistance beyond the risk associated with obesity and might improve the prediction of future metabolic risk.
3-Hydroxyisobutyrate
Amino acid
Aromatic amino acid
Adipose tissue insulin resistance
Branched-chain amino acid
Basic metabolic rate
Energy intake
Free fatty acids
Gestational diabetes mellitus
Impaired glucose tolerance
Normal glucose tolerance
Oral glucose tolerance test
Type 2 diabetes mellitus.
Data can be provided on request.
The authors declare no conflict of interest.
Agneta Holmäng designed the study; Ulrika Andersson-Hall, Carolina Gustavsson, and Louise Joelsson collected experimental samples and data; Anders Pedersen and Daniel Malmodin performed NMR analysis; and Ulrika Andersson-Hall and Agneta Holmäng interpreted the data and wrote the manuscript.
The authors thank Linda Rilby, RNM at Department of Obstetrics, Sahlgrenska University Hospital, for expert technical assistance. This work was supported by grants from the Swedish Research Council (12206), the Swedish Diabetes Association Research Foundation, and the Swedish Federal Government LUA/ALF agreement.
NMR data supporting annotation of methyl signal at 1.0679 ppm as 3-HIB. TOCSY cross peaks at 2.483, 3.535, and 3.703 ppm are in accordance with the 3-HIB HMDB entry 00023 values of 2.4727, 3.5244, and 3.6844. There is also a very weak signal in the natural abundance of 1H,13C-HSQC at a carbon chemical shift of 16.59 ppm, also in accordance with the 3-HIB methyl carbon shift in HMDB of 16.60 ppm.