Pregnant Chinese women were recruited from the Northwest Women and Children’s Hospital (NWCH) for participation in this case-control study. The participants were scheduled for delivery by caesarean-section from DEC. 2019 to MAR. 2020. Clinical information was collected from the 2^nd trimester study visit (24^th to 28^th weeks of gestation) until delivery. The 75-g, 2 h oral glucose tolerance test (OGTT) were performed at the 24^th – 28^th gestational weeks, and GDM was diagnosed based on the criteria of the International Association of Diabetes and Pregnancy Study Groups (17). The exclusion criteria was that: diabetes before pregnancy; fasting blood glucose (FBG) ≥ 7.0 mmol/L, 2-h glucose ≥ 11.1 mmol/L during OGTT; HbA1c ≥ 6.5% in the first trimester; eclampsia; young maternal age (less than 18); multiple pregnancy; other pre-existing diseases or gestational complications; and loss of vital data. Finally, from a total of 120 pregnant women, 57 subjects met the criteria and were included in further analysis.

Pre-pregnancy body mass index (pre-BMI), maternal age at delivery, and the gestational weight gain (GWG), gestational age at delivery and birth weight were noted. ISI composite index and the Stumvoll I index were adopted to evaluate the insulin sensitivity and insulin secretion, respectively. The information used for the calculation is obtained from the OGTT and the simultaneous insulin-release test (18–20). Women with normal glucose tolerance were defined as the control group (NGT, n = 13). We classified GDM subtypes by Powe’s definition (4), which is based on the distributions of the ISI composite index and the Stumvoll I index in the NGT group. GDM women with insulin sensitivity defect were defined if the ISI composite index was under the 25^th percentile of the range in the NGT group. GDM women with insulin secretion defect were defined if the Stumvoll I index was under the 25^th percentile of the range in the NGT group. Due to the limited number of participants in this study, we used the indexes of the women in the NGT group of our previous study (5) to calculate the normal range, since participants in both studies came from the same population over the same time periods. Therefore, we divided GDM women into the two subgroups: GDM with a primary defect of insulin-sensitivity (GDM-resistance, n = 17), GDM with a primary defect of insulin-secretion (GDM-dysfunction, n = 9). We excluded patients who had all the stated characters, or had the two indexes over the 25^th percentile.

The fasting blood samples were collected using the EDTA-coated tubes (Sarstedt, Newton, NC). At delivery, maternal blood was collected and centrifuged at 1000g for 15 minutes at 4°C. Then, the plasma was stored at -80°C for following assays. The vWAT specimen was obtained from the greater omentum (visceral) during caesarean-section, quickly frozen in liquid nitrogen and reserved at -80°C for following studies.

FBG was detected by the glucose oxidase approach (intra-assay variation factor was 2.1% and inter-assay variation factor was 2.6%). Plasma lipid profiles were detected using enzyme catalyzed approach according to the manufacturer’s procedure (A110-1-1, A111-1-1, A113-1-1, A112-1-1, Nanjing Jiancheng, China). Levels were quantified by a Microplate Reader with the wavelength of 546nm. The lipid levels included plasma triacylglycerol (TG), total cholesterol (CHO), low-density lipoprotein cholesterol (LDL-C) and high-density lipoprotein cholesterol (HDL-C). Plasma insulin levels were detected using commercial available kits (R-C-01-01, 5–180μU/mL, 3V Bioengineering, China). All lab tests were conducted in the certified lab of the NWCH with standard laboratory methods.

The plasma FGF21 levels at delivery were tested via a purchasable ELISA kit (CSB-E16844 h, Cusabio Biotech, Wuhan, China). All measurements were conducted according to the manufacturers’ protocol. The reference range of this detection was 15.6–1000 pg/mL with a susceptibility of 3.9 pg/mL. The coefficients of variability intra-assay and inter-assay were < 8% and < 10%, respectively.

(21)

GLU and INS measurements depicted as mmol/L and uU/mL, separately.

(20)

GLU and INS survey depicted as mmol/L and uU/mL, separately. We used the HOMA2-S and HOMA2-β at http://www.dtu.ox.ac.uk (11 Jan. 2016) as indexes to measure statuses of insulin sensitivity and insulin secretion.

The TRIzol reagent (Invitrogen, CA, USA) was employed to isolate total RNA from vWAT. 1μg total mRNA was reverse-transcribed into cDNA with the RT-PCR Kit (Thermo Scientific, USA). The entire qPCR was conducted with the iQ5 PCR thermocycler (Bio-Rad, USA). The primer sequences for the tested genes were presented in Supplementary Table 1 . The LightCycler protocol below was performed: 95°C for thirty seconds (pre-cultivation); 40 periods of 95°C for five seconds and 60°C for thirty seconds (amplify); and 81 periods of 55°C for ten seconds (melting curve). We included negative controls in the entire qPCR operations. The -△△Ct method was used to identify the comparative expressing scores. Each sample was analyzed in duplicate. Cyclophilin was used as the housekeeping gene. The efficiency of each primer was coherent within experiment groups.

The vWAT was homogenized with RIPA buffer (Beyotime, China) with the protease inhibitor and phosphatase inhibitor (Roche, Germany). Equal amount of protein was loaded in the 10% TGX stain-free gels (Bio-Rad) and was then transferred to PVDF membranes (Millipore). Images captured of stain-free gels were used to determine the total protein amount. The membranes were blocked with 5% non-fat dry milk and then were incubated with the primary antibodies at 4°C overnight. The Cell Signaling Technology offered antibodies below: phosphorylated and total protein kinase B [Akt (4723/4550)], phosphorylated and total extracellular signal-regulated kinase 1/2 [Erk1/2 (4695/4370)]. Then the membranes were treated with secondary antibodies at room temperature for 1h and processed for enhanced chemiluminescence detection. The ChemiDoc Touch Imaging System (Bio-Rad) was used to visualize the total protein quantity and the targeted protein signals. Image Lab software (Bio-Rad) was used to perform the densitometric assay of the total protein and the targeted protein signals in all lanes. The ratio of phosphorylated protein value to total protein value was employed to express the changes in protein activation.

Statistical analyses were done using SPSS 22.0 (SPSS Inc., USA). Data were displayed as means (SD or SEM) or median (IQR). One-way ANOVA was used for normal distribution data, the Kruskal–Wallis test was used for non-normal distribution continuous variables, and the Chi-squared test (or Fisher’s exact possibility test) was used for class variables to compare the differences across the three groups (the NGT group and the two GDM subgroups). Tukey’s test, Dunn’s test and Chi-squared test were carried out to conduct the pairwise comparisons between the NGT group and the experimental groups when the P-value from either of the above tests was < 0.05. The Bonferroni correction was used to modify the P-value for Chi-squared test paired contrasts. Regression model such as linear and multiple linear regression were applied to analysis the relationships of two continues variables.

Table 1 shows that women in the GDM-resistance group had higher pre-BMI (P = 0.006, Table 1 ) as compared with the NGT group. We observed no statistically significant differences in maternal age, GWG and infant birth weight among the three groups during pregnancy.

At the second trimester, compared with the women in the NGT group, women of the two GDM subtypes showed higher circulating levels of blood glucose during the OGTT test and larger glycemic area under the curve (AUC) (all P < 0.05). The insulin levels at all the time points and AUC (for insulin) exhibited a statistically significant increase (P < 0.001) in the GDM-resistance group when compared with the NGT group. By contrast, insulin level at the 2^nd hour and AUC (insulin) were decreased in the GDM-dysfunction group (P < 0.001). The levels of insulin secretion indicators (HOMA2-β, Stumvoll I index) were higher in the GDM-resistance group but lower in the GDM-dysfunction group when compared with the NGT group (all P < 0.05). However, the insulin sensitivity indicators [HOMA2-S (P < 0.05) and ISI composite index (P < 0.05)] were decreased in the GDM-resistance group when compared with the NGT group. No statistically significant differences were found in the insulin sensitivity indicators between the women in the GDM-dysfunction group and the NGT group.

Before delivery, compared with women in the NGT group, women in the GDM-resistance group had larger AUC (for glucose) (P = 0.035). However, the blood glucose levels showed no statistically significant differences at all of the time points during the OGTT test between the GDM-resistance group and the NGT group (all P > 0.05). By contrast, women in the GDM-dysfunction group showed higher blood glucose levels during the OGTT test and larger AUC (for glucose) when compared with the NGT group (all P < 0.05). The overall tendency of other OGTT related indicators (insulin levels during OGTT test, AUC for insulin, insulin secretion and sensitivity indicators) in women of all the groups were similar from the 2^nd trimester to delivery. Women in the GDM-resistance group exhibited lower level of plasma LDL-C (P = 0.033) as compared with the NGT group. Whereas, we found no statistically significant differences of plasma HDL-C, CHO and TG levels between women in the GDM-resistance group and the NGT group. Women in the GDM-dysfunction group had comparable lipid profile with the NGT group.

The circulating levels of plasma FGF21 in women of the GDM-dysfunction group were similar with the NGT group. However, women in the GDM-resistance group had elevated plasma FGF21 levels when compared with the NGT group (P < 0.001).

To investigate the alteration of insulin sensitivity and insulin secretion during pregnancy, we compared the ISI Composite Index and Stumvoll I Index between the second trimester and before delivery. The ISI composite index before delivery was improved when compared to that at the 2^nd trimester in the GDM-resistance group (P = 0.015, Figure 1A ). The ISI composite index at these two time points had no statistically significant differences in the GDM-dysfunction group and the NGT group. Meanwhile, we did not observe any statistically significant differences of the Stumoll I index between the 2^nd trimester and before delivery in each group ( Figure 1B ).

In order to find whether plasma FGF21 level is related with the clinical indicators, we performed linear correlation analyses. The plasma FGF21 levels were positively correlated with pre-BMI (r = 0.361, P = 0.028, Table 2 ) and negatively related to FBG (r = -0.718, P < 0.001, Table 2 ) and LDL-C (r = -0.438, P = 0.007, Table 2 ) at delivery. However, the plasma FGF21 levels were found negatively correlated with FBG (r = -0.574, P < 0.001, Table 3 ) and associated with the GDM-resistance group (r = 0.574, P < 0.001, Table 3 ) after adjusted by each other in the multiple linear regression. Plasma FGF21 concentrations had no linear relationships with GWG, CHO, TG, HDL-C.

In order to further study the FGF21 signaling in target organs, we detected the relative gene and protein expression of FGF21 receptors and downstream signaling pathways in vWAT. Compared with the NGT group, gene expression of fibroblast growth factor receptor 1c (FGFR1c) was increased (P < 0.05, Figure 2A ), while gene expression of β-Klotho (KLB) was decreased (P < 0.05) in the GDM-resistance group ( Figure 2B ).

Furthermore, we found no statistically significant differences in the phosphorylation level of Akt (Ser 473) among the three groups ( Figure 2C ). However, the phosphorylation level of Erk1/2 was increased in the GDM-resistance group (P < 0.05, Figure 2D ) when compared with the NGT group. The Akt and Erk1/2 signaling was not significantly altered in the vWAT of GDM-dysfunction group as compared with the NGT group.

We found higher mRNA expression of glucose transporter-1 (GLUT1), insulin receptor-β (IR-B) and peroxisome proliferators activated receptor-γ (PPAR-γ) (all P < 0.05, Figures 3A, C, D ) in the vWAT of the GDM-resistance group when compared with that in the NGT group. The mRNA expression of glucose transporter-4 (GLUT4), adiponectin, C1Q and collagen domain containing (ADIPOQ), adipose triglyceride lipase (ATGL) and phosphoprotein perilipin (PLIN-1) of the GDM-resistance group showed no statistically significant differences as compared with the NGT group ( Figures 3B, E–G ). Meanwhile, in the GDM-dysfunction group, we observed no statistically significant differences in the mRNA expression of GLUT1, GLUT4, ADIPOQ, ATGL, PLIN-1, IR-B, and PPAR-γ when compared with the NGT group. In addition, we hardly detected the mRNA expression of FGF21 in vWAT (data not shown).

In Table 4 , circulating FGF21 levels were positively correlated with the relative gene expression of GLUT1 (r = 0.383, P = 0.021), IR-B (r = 0.355, P = 0.042) and PPAR-γ (r = 0.402, P = 0.015). Plasma FGF21 concentrations at delivery had no linear relationships with the relative gene expression of GLUT4, ADIPOQ, ATGL and PLIN-1.