Proteome profiling of gestational diabetes mellitus at 16‐18 weeks revealed by LC‐MS/MS

Abstract Background The practices used to diagnose gestational diabetes mellitus (GDM) could only be carried out around the time of detectable symptoms, and predictive capacity is little. Methods LC‐MS/MS was conducted to explore overview proteomics for GDM complicated pregnant woman at 16‐18 gestation weeks, while normal pregnant for control. Enzyme‐linked immunosorbent assay was further applied in an independent cohort of 15 GDM cases and 15 controls for verification. Results The results indicated that 24 protein expression levels were significantly changed in GDM group samples, and inflammation, oxidative stress, insulin resistance, blood coagulation, and lipid homeostasis were associated with GDM. The abnormal expression of CRP and IGFBP2 was verified in the first‐trimester maternal plasma in women who subsequently developed GDM. Conclusions This study not only identified 24 potential predictive biomarkers for GDM also provided a global overview of protein rearrangements induced by GDM.

capacity is little. At 24-28 gestation weeks, when GDM diagnosis carried out, harms may have potentially occurred to both mother and child. 10 Effective intervention and management could positively affect maternal and fetal outcomes; therefore, predictive biomarker exploration is desirable.
Previous studies investigated the potential value of first-trimester maternal serum markers of GDM reported promising results, such as insulin resistance (sex hormone-binding globulin and homeostasis model assessment index), inflammation (high sensitive C-reactive protein), and adipocytokines (adiponectin, visfatin, and leptin) which have been measured in the first or second trimester of GDM. [11][12][13][14][15] Mass spectrometry-based proteomics of human plasma could provide a more profound understanding in the pathogenesis of many diseases, such as hypertension, 16 cancer, 17 GDM, 18 and major depressive disorder. 19 The above studies were based on that proteins secreted into human body fluid could reflect disease states, and alterations in expression level are indicative of developing lesion. However, due to the complexity and high dynamic variation of the human plasma components, the overview proteomics of GDM remain challenging.
The aim of this study was to explore the proteomics of GDM for further understanding the subsequent development of gestational diabetes. In our study, protein expression levels and global correlation analysis including inflammation, oxidative stress, insulin resistance, blood coagulation, and lipid homeostasis were investigated in GDM and normal pregnancy by liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis.

| Patient demography
Clinical characteristics of both the GDM and control samples are shown in Table 1. The mothers with GDM were slightly older and had higher pre-pregnancy BMI values when compared with the control group. There was no significant difference in gestational weeks between the two groups. OGTT of fast, 1, and 2 hours was all significantly different.

| Chemical and reagents
Urea, dithiothreitol (DTT), iodoacetamide (IAA), ammonium bicarbonate, trifluoroacetic acid (TFA), and SOLA HRP 96-well SPE plate were bought from Thermo Fisher (Thermo Fisher Scientific). Trypsin was bought from Promega (Promega). Lysyl endopeptidase (Lys-C) was bought from Wako (Wako). All organic reagents used in the experiment are chromatographically pure reagents. The experimental water is Milli-Q ultrapure water (Millipore). Other reagents are analytically pure reagents unless otherwise indicated.  Healthy controls were selected based on similar gestational weeks and gestational age to maintain similar maternal baseline characteristics. Subjects who displayed a history of type-2 diabetes were excluded, as were smokers or those with a chemical dependency, and those who exhibited fetal congenital anomalies or any other confounding pathology (including hyperthyroidism, hypothyroidism, hypertension, and hyperlipidemia). The approval of this study was granted by Chinese PLA General Hospital Ethics Committee and met the Declaration of Helsinki. then washed with 0.1% trifluoroacetic acid (TFA) and finally eluted with 30% acetonitrile (ACN) and 60% ACN containing 0.1% TFA.

| Plasma protein digestion
Elute was lyophilized in a vacuum centrifuge for LC-MS/MS proteome analysis.

| Data analysis
All DIA files were extracted from home generated plasma library containing 73 466 precursors, 55 433 modified peptides and 5190 proteins using Spectronaut X (Biognosys). Raw MS data files were converted to HTRMS files with HTRMS converter (Biognosys).
HTRMS files were imported to Spectronaut with default parameters with the decoy generation set to "mutated." Cutoff of fold change > 1.3 or <0.3785 and Q < 0.05 (FDR-corrected P value for t test) was set for differentially expressed proteins. Correlation and ROC analysis were carried out using R software version 3.5.3.

| Bioinformatics analysis
Gene Ontology (GO) analysis and UniProt-KB keyword analysis for the significantly changed proteins and correlated proteins were conducted on the Database for Annotation, Visualization and Integrated Discovery (DAVID) version 6.8. 20,21 Fisher's exact test was used in determining the significant enrichment terms, and .05 was set as the threshold (P-values). Only significantly changed category terms were reported in this study.
Inflammation system proteins included proteins with Uniprot-UK keywords for inflammatory response, immunity, innate immunity, complement pathway, complement alternate pathway, acute-phase, membrane attack complex, cytolysis, and antimicrobial.

| Global profiling of plasma proteins in GDM and control samples
To  Table S1. Ranked proteins in control and GDM group indicated similar distributions in plasma proteome ( Figure 1C,D).

| GDM induced plasma proteome rearranges
Our aim was to reveal potential biomarkers to predict GDM at early second-trimester through plasma proteomic analysis. The study design is illustrated in Figure 2A. Through fold change and t test analysis, 24 proteins were found to change significantly ( Figure 2B and  We annotated the 24 significantly changed proteins with gene ontology (GO). GOCC (cell component) indicated that these proteins located in extracellular, blood microparticle, and cell surface (Table S3). Significantly enriched GOBP (biological process) included acute-phase response, female pregnancy, cellular proteins metabolic process, and cellular response to hormone stimulus ( Figure 2C). And based on Fisher's exact test, 4 keywords were obtained (P ≤ .5) for GOMF (molecular function) including complement component C1q binding, virion binding, endopeptidase inhibitor activity, and protein homodimerization activity (Table S3). As the complicated relationship of these proteins, we set out to exploit global co-regulation proteins.

| Changed proteins have diagnostic value
To examine the potential diagnostic value of the changed proteins, receiver operating characteristic (ROC) analysis was performed.
ROC curve was constructed between OGTT diagnose results of subjects and binary logistic regression predictions on the two-category attribute based on the 24 changed proteins. The area under the curve (AUC) was 1 ( Figure 3) and indicated these proteins as a perfect classifier.

| Global correlation maps reveal protein coregulations
Proteins work together in a complex network in biosystems, and the concentrations and activities of proteins in the network are rigorously controlled. Therefore, after comparison of GDM and control samples, we investigated the co-regulated proteins to explore pathological mechanism of GDM. At least 50% proteins that quantified in samples were included. Pairwise correlations between these proteins and four clinical parameters were calculated to construct global correlation heatmap ( Figure 4A). In the correlation heatmap, proteins and clinical parameters were grouped with hierarchical cluster analysis and correlation coefficients were coded as colors.
The global correlation map was constructed with 475 proteins, faster OGTT, 1 hour OGTT, 2 hours OGTT, and BMI values. Protein abundant levels, OGTT results, and BMI generated a matrix of 114 481 correlation coefficients ( Figure 4A and Table S4). As previous studies indicated that proteins and clinical parameters associated with the same underlying regulatory mechanism would cluster in the same area, we used this map to explore GDM induced co-regulated proteins and clinical parameters. Correlation network of the Uniprot-UK keyword annotated proteins in the five co-regulated areas revealed the connections between these proteins. Only proteins with absolute correlation coefficients values above 0.5 were selected to display their relationships ( Figure 4B and Table S5). This network highlighted the connections of the main physiological processes. Strong correlations were observed within inflammation system proteins, and the correlations of inflammation with blood coagulation, lipid homeostasis, and membrane proteins were also captured. Two antioxidative stress protein (P08294, P27169) showed correlation with blood coagulation, inflammation system and lipid homeostasis. Acetylation proteins only showed correlations within these proteins and did not show correlation with the other proteins.

| Protein expression verificated by ELISA
Consistent with the LC-MS/MS data, ELISA analysis verified the abnormal expression of CRP and IGFBP2 in the first-trimester maternal plasma in women who subsequently developed GDM ( Figure 5). Both proteins were examined for their performance in differentiating between GDM and control samples. The area under the curve (AUC) obtained for CRP was 0.953 and for IGFBP2 1.0 at P < .001 ( Figure 6).

| D ISCUSS I ON
Gestational diabetes mellitus could alter human metabolism, and the potential molecular mechanisms of GDM remains unknown.  Table S6).
The scattered distribution of these proteins in Gene Ontology biological processes may indicate the mildness of proteomic rearrangements induced by GDM at early second trimester (16)(17)(18) gestation weeks). Oxidative stress has been shown to interact with inflammation to regulate disease outcomes. 23 The changed oxidative stress metabolite biomarkers such as 8-isopeostane 24 and malondialdehyde, 25 plasma oxidation/redox status-related proteins such as protein carbonyl and nitrotyrosine, 26 antioxidant activity enzymes such as para-oxonase1 (PON1) 26  Blood coagulation and lipid homeostasis were also documented to be involved in pregnancy complicated with GDM.
Hypercoagulability was observed to be enhanced in GDM compared with normal pregnancy group. In precious study, GDM is observed to associate with shortened activated partial thromboplastin time (APTT), increased activity of antithrombin III (ATIII), and a higher plasminogen activator inhibitor (PAI-1) content levels. 30 This study may reveal that plasma hypercoagulability in GDM pregnant women was induced by co-regulations of many blood coagulation proteins.
Lipid homeostasis is involved in pregnancy complicated with GDM and whether lipids and lipoproteins could be used as biomarker to predict GDM need further study for the contradictory results.
An early study indicated that concentrations of very-low-density, low-density and high-density lipoproteins, plasma cholesterol, and triglyceride changed along gestation time during pregnancy. 31 The study on Japanese women at 20-28 weeks of gestation indi- shown low circulating IGFBP-2 is associated with reduced insulin sensitivity. Our study showed that lower expression of IGFBP-2 in GDM group compared with control by LC-MS/MS, which also identified by ELISA. Therefore, our results presented that IGFBP-2 could be used as the early predictive biomarker for GDM prenatal screening.
However, the limited number of cases in present study may result in an overestimation. Further research is required by using a larger population representative of GDM to validate the data obtained in early predicting GDM. Abnormal expression of CRP and IGRBP2 in the first-trimester maternal plasma of GDM women was identified by ELISA. These proteins displayed great potential to be acted as early predictors for GDM prenatal screening.

| CON CLUS IONS
In summary, our study explored 24 potential predictive biomarkers for GDM and provided a global overview of protein rearrangements induced by GDM at early second trimester. With global correlation analysis, the co-regulations of inflammation, oxidative stress, insulin resistance, blood coagulation, and lipid homeostasis were also revealed in GDM development. The abnormal expression of CRP and IGFBP2 was verified by ELISA in the first-trimester maternal plasma in GDM women.

ACK N OWLED G M ENTS
The authors acknowledge the efforts and contributions of all the participating volunteers.