Using metabolomics and proteomics to identify the potential urine biomarkers for prediction and diagnosis of gestational diabetes

Summary

Gestational diabetes mellitus (GDM) is one of the most common metabolic complications during pregnancy, threatening both maternal and fetal health. Prediction and diagnosis of GDM is not unified. Finding effective biomarkers for GDM is particularly important for achieving early prediction, accurate diagnosis and timely intervention. Urine, due to its accessibility in large quantities, noninvasive collection and easy preparation, has become a good sample for biomarker identification. In recent years, a number of studies using metabolomics and proteomics approaches have identified differential expressed urine metabolites and proteins in GDM patients. In this review, we summarized these potential urine biomarkers for GDM prediction and diagnosis and elucidated their role in development of GDM.

Search strategy and selection criteria.

Data for this review were identified by searches of PubMed and references from relevant articles using the search terms: “gestational diabetes mellitus” and “urine” cross-referenced with “biomarker”, “prediction”, “diagnosis”, “metabolite”, “metabolomics”, “lipid”, “protein” and “proteomics”. Only articles published in English up to Oct 1, 2023 were included.

Introduction

Gestational diabetes mellitus (GDM) is a type of diabetes diagnosed in the second or third trimester of pregnancy that is not clearly overt diabetes before gestation.¹ With a steady increase in global morbidity observed over the last four decades, GDM has become one of the most common metabolic complications during pregnancy.² Nine to twenty-six percent of pregnant women are diagnosed with GDM based on different criteria.³ Epidemiological studies have identified some risk factors for GDM, including ethnicity, increased maternal age, overweight/obesity, previous gestational diabetes history and a family history of type 2 diabetes mellitus (T2DM).⁴ GDM can raise the risk of adverse pregnancy and neonatal consequences, including spontaneous abortion, congenital anomalies, preeclampsia, perinatal asphyxia, respiratory distress, fetal demise, macrosomia, cardiomyopathy and hyperbilirubinemia.⁵ Although recognized and treated during pregnancy, GDM has a life-long impact on maternal and fetal health. Women with a history of GDM are more likely to develop T2DM, cardiovascular disease (CVD) and chronic kidney disease (CKD) later in life, and their offspring tend to have metabolic diseases, such as diabetes, hypertension and obesity.⁶

Currently, diagnostic criteria for GDM are unified between obstetric and diabetes organizations from different countries.7, 8, 9, 10 The one-step and two-step strategies, both implemented at 24–28 weeks of gestation, are widely used and accepted internationally.¹ Nevertheless, current criteria for diagnosing GDM on the basis of plasma glucose levels fail to identify all GDM patients and cannot distinguish pregnant women who may develop postpartum diabetes or CVD.¹¹ Overall, early interventions for GDM, such as physical activity, diet intervention and probiotic supplementation, have been proven effective in reducing GDM risk, as confirmed by several randomized controlled trials (RCTs).12, 13, 14 Consequently, it is essential for early prediction, precise diagnosis and timely intervention of GDM, which not only protects mothers and offspring against disadvantageous influences but also decreases health care use and cost.

The mechanism of GDM is typically described as increased insulin resistance (IR) and pancreatic β-cell defects.¹⁵ Appearance of GDM is a gradual process. Biomarkers act as a reliable tool to reflect the internal condition of an organism and are utilized to evaluate the risk of disease development and response to therapy.¹⁶ Numerous studies about GDM biomarkers have been published, mainly focusing on maternal blood biomarkers.¹⁷ Moreover, some insightful researchers have investigated urine biomarkers because urine measurement has several advantages in clinical applications. For example, urine is accessible in large quantities, can be collected noninvasively, provides an abundant ingredient profile to analyze, and has the capacity to reflect biochemical imbalances in the organism. Compared with other biofluids, handling of urine samples is also relatively simple.¹⁸

In the last decade, many studies have been conducted to discover potential urine biomarkers of GDM with the advancement in metabolomics and proteomics techniques. The aim of the studies focusing on the first and early second trimesters of pregnancy is to identify early predictive molecules, and the aim of the studies focusing on the second and third trimesters of pregnancy is to identify urinary biomarkers that are indicative of the diagnosis of GDM and maternal and fetal outcomes. Metabolites and proteins in urine have been confirmed to vary in GDM patients at different gestational weeks (Table 1). This review aimed to objectively and thoroughly summarize studies on the application of metabolomics and proteomics approaches in GDM, highlighting aspects of potential urine biomarkers for prediction and diagnosis. We also compared the sensitivity and specificity of different predictive and diagnostic models (Table 2), clarified the possible roles of urine metabolites and proteins in the pathogenesis of GDM, and proposed strategies for future metabolomic and proteomic studies of GDM.

Table 1.

Summary of studies discovering altered urine components in GDM patients.

Authors	Countries	Techniques	Study design	Gestational weeks	Changes in urine
Miriam Leitner et al.19	Austria	SID-MS	cross-sectional study; GDM (n = 14) vs NGT (n = 18)	12∼26 weeks	↑: l-tryptophan, serotonin, N-acetylserotonin, 5-HIAA; ↓: 5-methoxytryptamine, melatonin
Kai P. Law et al.20	China	UPLC-MS	longitudinal cohort study; GDM (n = 27) vs NGT (n = 34)	11∼14 weeks, 23∼27 weeks and 29∼33 weeks	↑: indoleacetaldehyde, serotonin, 5-hydroxykynurenamine, xanthine, 1-methyladenosine, 1-methylhypoxanthine and N4-acetylcytidine (in the 1st-3rd trimesters); indoleacetic acid, indole-3-acetamide, oxitriptan and hypoxanthine (in the 1st and 2nd trimesters); uric acid and 7-methylguanine (in the 1st and 3rd trimesters); xanthosine (in the 2nd trimester)
Danuta Dudzik et al.21	Poland	CE-TOF/MS	cross-sectional study; GDM (n = 20) vs NGT (n = 20)	22∼28 weeks	↑: hystidine, glutamine, phenylalanine, tryptophan and cystine; ↓: carnitine
Yamilé López-Hernández et al.22	Mexico	UPLC-MS	cross-sectional study; GDM (n = 24) vs NGT (n = 11)	third trimester	↑: 5-carboxy-alpha-chromanol, aspartame, DG (24:0/14:1), l-tryptophan, l-urobilinogen, Cer (d18:0/23:0), SM (d18:0/22:0), 11-oxo-androsterone-glucuronide, cortolone-3-glucuronide, tetrahydroaldosterone-3-glucuronide, 5-androstene-3b,16b,17a-triol, 21-deoxycortisol, 11b,17a,21-triydroxypreg-nonolone and cucurbitacin c
Xing Wang et al.23	China	GC/MS	cross-sectional study; GDM (n = 59) vs NGT (n = 48)	N/A	↑: d-galactose, 3-methoxytyrosine and d-glucose; ↓: 2-O-methyl-l-ascorbic acid, citraconic acid, N-methylglutamic acid, cysteine-glycine, glutamine, glycine, galactitol, trehalose, isocitric acid, 2-hydroxyhexanoic acid, nicotinic acid and 3-dehydroshikimic acid
Chunfang Qiu et al.24	USA	LC-MS/MS	nested case-control study; GDM (n = 25) vs NGT (n = 25)	16∼17 weeks	↑: ethylmalonate and pyruvate; ↓: adipate
Kenichi Sakurai et al.25	Japan	HILIC-MS/MS	nested case-control study; GDM (n = 36) vs NGT (n = 31)	first and early second trimesters	↑: 1,3-diphosphoglycerate, shikimate-3-phosphate and N-acetyl-l-alanine; ↓: ethanolamine and methionine
Chunfang Qiu et al.26	USA	ELISA	nested case-control study; GDM (n = 55) vs NGT (n = 43)	around 16 weeks	↑: 8-OHdG
Mehmet Oğuz Erbağcı et al.27	Japan	LC-MS/MS	longitudinal cohort study; GDM (n = 33) vs NGT (n = 84)	11∼14 weeks	↑: 8-OHdG
Wenjin Fu et al.28	China	ELISA	cross-sectional study; GDM (n = 101) vs normal pregnant women (n = 20) vs nonpregnant women (n = 24)	N/A	↑: L-FABP (GDM > normal pregnant women > nonpregnant women)
Zhiying Hu et al.29	China	MALDI-TOF/MS	case-control study; GDM (n = 74) vs NGT (n = 31)	second and third trimesters	↑: ITIH4
Zhiying Hu et al.30	China	MALDI-TOF/MS	case-control study; GDM (n = 60) vs normal pregnant women (n = 31) vs nonpregnant women (n = 31)	second and third trimesters	↑: coagulation factor IX, TBC1D5a and immunoglobulin kappa constant (GDM > normal pregnant women and nonpregnant women)
Brian J. KOOS et al.31	USA	UPLC-MS/MS	nested case-control study; GDM (n = 46) vs NGT (n = 46)	around 12 weeks	↑: argininate, saccharopine, nicotinate ribonucleoside, 7,8-dihydroneopterin and phenol glucuronide
Joana Pinto et al.32	Portugal	NMR	case-control study; non-treated GDM (n = 18) vs NGT (n = 34)	second and third trimesters	↑: choline, creatine, galactose and threonine; ↓: hippurate, 3-HIVA, lysine and PAG

Abbreviation: GDM: gestational diabetes mellitus; SID-MS: stable isotope dilution direct infusion electrospray ionization mass spectrometry; NGT: normal glucose tolerance; 5-HIAA: 5-hydroxyindoleacetic acid; UPLC-MS: ultra-performance liquid chromatography-tandem mass spectrometry; CE-TOF: capillary electrophoresis time-of-flight; MS: mass spectrometry; DG: diglyceride; Cer: ceramide; SM: sphingomyelin; GC: gas chromatography; LC: liquid chromatography; HILIC-MS: hydrophilic interaction chromatography tandem mass spectrometry; 8-OHdG: 8-hydroxy-2-deoxyguanosine; L-FABP: liver-type fatty acid-binding protein; MALDI-TOF: matrix-assisted laser desorption ionization time-of-flight; ITIH4: Inter-alpha-trypsin inhibitor heavy chain H4; TBC1D5a: TBC1 family member 5 isoform a; NMR: nuclear magnetic resonance; 3-HIVA: 3-hydroxyisovalericacid; PAG: phenylacetylglutamine.

Table 2.

Urine biomarker models for prediction or diagnosis of GDM.

Models	Study design	Gestational weeks	Sensitivity	Specificity	The AUC value
Urine serotonin metabolites (l-tryptophan, serotonin, 5-HIAA, melatonin and 6-hydroxymelatonin) and 11 plasma metabolites19	cross-sectional study; GDM (n = 14) vs NGT (n = 18)	12∼26 weeks	N/A	N/A	0.990
Urine metabolites (trehalose and 3-dehydroshikimic acid), bacteria (Enterobacteriaceae and Enterococcaceae) and fecal metabolites (5-hydroxyindoleacetic acid and valine)23	cross-sectional study; GDM (n-59) vs NGT (n = 48)	N/A	N/A	N/A	0.957
Urine metabolites (dihydroorotate, argininate, phenylglucuronide, 7,8-dihydroneopterin, nicotinate ribonucleoside, saccharopine and lanthionine)31	nested case-control study; GDM (n = 46) vs NGT (n = 46)	around 12 weeks	97.8%	95.7%	0.993
Urine proteins (coagulation factor IX, TBC1D5a and immunoglobulin kappa constant), FPG and HbA1c30	case-control study; GDM (n = 60) vs normal pregnant women (n = 31) vs nonpregnant women (n = 31)	second and third trimesters	76.7%	64.5%	0.758

Abbreviation: GDM: gestational diabetes mellitus; AUC: area under the curve; 5-HIAA: 5-hydroxyindoleacetic acid; TBC1D5a: TBC1 family member 5 isoform a; FPG: fasting plasma glucose.

Urine metabolomics

Approximately 4500 metabolites have been identified in urine in relation to approximately 600 human conditions.³³ The potential ability of urine metabolites to predict and diagnose diseases has been extensively researched, including but not limited to cancers,³⁴ kidney diseases³⁵ and metabolic diseases.³⁶ Metabolomics is the qualitative or quantitative analysis of a variety of small molecule metabolites that are intermediate or end products of all metabolic pathways in a living organism.³⁷ Metabolomics analysis strategies are generally categorized as untargeted and targeted. Untargeted methods are more appropriate for exploring candidate biomarkers or metabolic mechanisms in diseases when it is unclear which metabolites or metabolic pathways are crucial to the research question. In contrast, targeted methods are typically utilized for verification of biomarkers and investigation of specific biological pathways by analyzing metabolites with similar physicochemical properties or those participating in the same biochemical pathways.³⁷ Liquid chromatography (LC), gas chromatography (GC) and capillary electrophoresis (CE) coupled with mass spectrometry (MS) or proton nuclear magnetic resonance (NMR) are the most frequently employed platforms for metabolomics. MS has the advantages of high sensitivity, high resolution, low sample consumption, fast analysis speed and a wide range of data coverage; its disadvantages include complex sample preparation, requirement for different chromatographic techniques and poor repeatability of data.³⁸ NMR does not require chromatographic processing and sample preparation is simple. It can detect tissue samples in situ and provide accurate structural information about metabolites. However, its shortcomings are low sensitivity, incomplete metabolite coverage and higher detection cost. Since both NMR and MS have advantages and disadvantages and are highly complementary, the combination of these two techniques can improve the overall quality of the study.³⁷ The complex data produced by high-throughput metabolomics require advanced algorithms to process, including traditional approaches such as univariate analysis and logistic regression, as well as novel approaches such as machine learning.³⁹ Changes in metabolic composition directly reflect disease processes, environmental exposures, and toxicological and nutritional intakes at a certain time since metabolites act as substrates or products in metabolic pathways. After validation, these key metabolites can be applied for disease diagnosis, treatment evaluation, and even prediction of disease susceptibility.⁴⁰

GDM is influenced by multiple-level factors, which involve various metabolic pathway disorders, such as amino acids, lipids and purines.⁴¹ Due to its unparalleled capability to measure thousands of metabolites from intricate biological systems, metabolomics is becoming a widely utilized technique for biomarker discovery and in the field of GDM research.⁴² Metabolites in urine have the potential to become biomarkers for GDM and provide further insight into the etiology and pathophysiology of GDM.

Amino acid metabolites

Growing evidence indicates a close association between amino acid metabolites and metabolic disorders, such as obesity, IR and T2DM.43, 44, 45 With advancement of metabolomics, an increasing number of amino acid metabolites related to GDM have been detected, especially aromatic amino acids (AAAs) and branched-chain amino acids (BCAAs).

AAA

AAAs are amino acids possessing a phenyl ring structure that contains tyrosine, phenylalanine and tryptophan (Trp). Several studies have shown that there is a positive correlation between the circulating (serum or plasma) level of AAAs and development of GDM.³⁸ Tryptophan is an essential amino acid obtained from diet. There are three main pathways of tryptophan catabolism: (1) tryptophan can be metabolized to serotonin via tryptophan hydroxylase (TPH) and aromatic amino acid decarboxylase (AADC); (2) tryptophan can enter the tryptophan-kynurenine (Trp-KYN) pathway, which is regulated by indoleamine 2,3-dioxygenase 1 (IDO1) and tryptophan 2,3-dioxygenase (TDO); (3) tryptophan can be directly converted by the gut microbiota into several molecules, including indole and its derivatives, which are ligands of the aryl hydrocarbon receptor (AhR).^46,47 Trp metabolism can affect the function of islet β-cells, IR, intestinal barrier and angiogenesis in diabetes mellitus.⁴⁸ Notably, multiple studies have found that urine tryptophan catabolism is aberrant in GDM patients, which is worthy of specific discussion.19, 20, 21, 22

Serotonin (5-hydroxytryptamine, 5-HT) is synthesized from tryptophan through a series of enzymatic processes. It regulates various functions, including but not limited to the well-known roles in emotion and feeding behaviors, as well as its influence on metabolic homeostasis.⁴⁹ A study by Miriam Leitner et al. employed untargeted GC-MS identification of plasma metabolites and targeted stable isotope dilution direct infusion electrospray ionization mass spectrometry (SID-MS) detection of urine metabolites of the tryptophan-melatonin-serotonin pathway.¹⁹ Participants included fourteen women with GDM and eighteen women with normal glucose tolerance at 12∼26 weeks of pregnancy, and plasma metabolites such as tryptophan significantly distinguished the GDM patients from healthy controls. The urine target analytes l-tryptophan, serotonin, N-acetylserotonin, and 5-hydroxyindoleacetic acid (5-HIAA) were increased and 5-methoxytryptamine and melatonin reduced in the GDM patients in comparison to the healthy pregnant women. After combining the data on urine serotonin metabolites (l-tryptophan, serotonin, 5-HIAA, melatonin and 6-hydroxymelatonin), the accuracy of the GDM predictive model was enhanced, with an area under the curve (AUC) value of 0.99.¹⁹ In mouse models, evidence suggests that serotonin synthesis in pancreatic islets activated by lactogenic signaling during pregnancy stimulates pancreatic β-cell proliferation.⁵⁰ Serotonin within pancreatic β-cells induces insulin secretion through serotonylation, whereby serotonin is covalently coupled with the small GTPases Rab3a and Rab27a.⁵¹ These findings support the hypothesis that elevated serotonin levels potentially function as a compensatory mechanism for sustaining standard glycemia in the event of a disturbed metabolic state, such as in GDM.¹⁹ Accordingly, abnormalities in tryptophan and serotonin metabolism may be fundamental in the pathogenesis of GDM, which requires confirmation by further studies analyzing larger cohorts.

Indoleamine 2,3-dioxygenase 1 (IDO1), the rate-limiting enzyme of the Trp-KYN pathway, has been confirmed to cause dysregulation of Trp-KYN pathway and mediate pregnancy IR. K.P. Law et al. investigated the dynamic change of the urine metabolome in a longitudinal cohort of thirty-four healthy women and twenty-seven women diagnosed with GDM. The researchers conducted ultra-performance liquid chromatography-tandem mass spectrometry (UPLC-MS) analysis on urine samples collected at 11∼14 weeks, 23∼27 weeks and 29∼33 weeks of gestation. The trajectories of the urine metabolome illustrated that the majority of tryptophan metabolites were upregulated in women with GDM at all stages of gestation, especially the metabolites in the Trp-KYN pathway. 5-Hydroxykynurenamine, serotonin, indoleacetic acid, indoleacetaldehyde, oxitriptan and indole-3-acetamide demonstrated good potential as GDM biomarkers, with an AUC value > 0.7 in the first and second trimesters.²⁰ A conventional view of the pathogenesis of GDM is that hormones secreted by the placenta act as the dominant factor for IR in GDM. In contrast, this study found that activation of the Trp-KYN pathway in GDM occurred before the physiological effects of placental hormones were exerted, offering support for the hypothesis that GDM may be a predisposed condition in which a prediabetic state is completely expressed under stress due to alteration of metabolism in pregnancy. Chronic low-grade inflammation in women with GDM activates IDO1, which upregulates the Trp-KYN pathway and ultimately mediates IR.⁴⁷ Genetic knockout or pharmacological inhibition of IDO1 in pregnant female mice demonstrate a reduced kynurenine to tryptophan ratio (K/T), decreased intestinal inflammation and improved IR.⁵² Catabolism of tryptophan along the Trp-KYN pathway represents a promising opportunity for management of IR and GDM.⁵³

BCAA

BCAAs are amino acids with a branched side chain structure consisting of leucine, valine and isoleucine. These amino acids are crucial for maintaining energy homeostasis, nutritional metabolism and immunity.⁵⁴ The positive relationship between the circulating (serum or plasma) levels of BCAAs and the risk of IR, T2DM and GDM has been reported by many studies in recent years.⁵⁵ One possible mechanism is that elevated levels of BCAAs activate mammalian targets of rapamycin complex 1 (mTORC1), ultimately contributing to IR via phosphorylation of insulin receptor substrate 1 (IRS-1).⁵⁶ Another mechanism is that toxic BCAA metabolites accumulate, leading to oxidative stress, mitochondrial malfunction and ultimately impaired insulin function.⁵⁷ The abundance of valine in the blood of GDM women is increased,58, 59, 60 and similar results were obtained for valine in feces in Wang's study. Considering that the gut microbiota can regulate host metabolic homeostasis, this study explored interaction between the gut microbiota and relevant stool and urine metabolites in GDM.²³ Compared to healthy controls, three upregulated metabolites and twelve downregulated metabolites were observed in the urine of GDM subjects. These urinary metabolites are products of amino acid and carbohydrate metabolism according to functional clustering analysis. Maternal microbial composition was altered in feces from GDM patients, as characterized by increased abundance of the family Lachnospiraceae and decreased abundance of the families Enterobacteriaceae and Ruminococcaceae. Importantly, Lachnospiraceae and Enterobacteriaceae showed significant co-occurring relationships with metabolites related to amino acid and carbohydrate metabolism in GDM subjects. Hence, a novel combinatorial marker panel including bacteria (Enterobacteriaceae and Enterococcaceae), urine metabolites (trehalose and 3-dehydroshikimic acid) and fecal metabolites (5-hydroxyindoleacetic acid and valine) effectively distinguished GDM from healthy women, with an AUC value of 0.957. The strategy of integrating metagenomics and metabolomics provides a new perspective for understanding the pathogenesis and pathophysiology of GDM that alteration of gut microbiota may initiate GDM by disrupting the host's amino acid and carbohydrate metabolism.²³

Lipid metabolites

Lipids, which are divided into eight principal categories, fatty acids (FAs), glycerolipids, glycerophospholipids, sphingolipids, sterol lipids, prenol lipids, saccharolipids and polyketides, play a significant part in energy metabolism and cell signaling.⁶¹ They also constitute crucial elements of hormones, cell membranes and lipid particles. Lipid homeostasis represents a unique readout for health conditions due to the dependence of cell and organ functions on lipids. Dysfunction of lipid metabolism can initiate or aggravate a variety of diseases, including atherosclerosis and stroke, diabetes and obesity, neurodegenerative and neurological disorders, autoimmune disorders, and even cancers.⁶² Lipidomics, a branch of metabolomics, differs from general metabolomics because of the particular physicochemical properties of lipids, contrary to other water-soluble cellular metabolites.⁶³ Some lipid metabolites are strongly associated with the risk of T2DM, extending the current spectrum of predictors of T2DM in normoglycemic populations.⁶⁴ There are also some lipid metabolites that can be used as biomarkers of diabetes complications, including cardiovascular events⁶⁵ and polyneuropathy.⁶⁶ The main lipid differences closely associated with GDM are FAs, phospholipids, glycerolipids, glycerophospholipids, sphingolipids and steroids.67, 68, 69 Due to their complex structural composition and multiple functionalities, urine lipid metabolites of GDM will be discussed separately below.

FA

FAs are the simplest lipid molecules and act as regulators of membrane properties, energy suppliers and storage materials.⁷⁰ FAs can be classified based on the saturation degree of the hydrocarbon chain into three groups: saturated fatty acids (SFAs), monounsaturated fatty acids (MUFAs), and polyunsaturated fatty acids (PUFAs).⁶¹ Qiu et al. utilized LC-MS/MS technology to measure variation in organic acid biomarkers in urine samples collected on average at 16 weeks of pregnancy. These organic acid metabolites are characterized by altered fatty acid metabolism (ethylmalonate and adipate) and carbohydrate metabolism (pyruvate). The risk of subsequent GDM correlated positively with the concentration of ethylmalonate and negatively with the concentration of adipate in maternal urine during early pregnancy.²⁴ The profile of organic acids in urine can provide important clues to maternal fatty acid and carbohydrate metabolism during pregnancy and may help to understand the pathogenesis of GDM. Total plasma SFAs levels are higher before the onset of GDM. Evidence from in vitro and animal models demonstrate that SFAs suppress the insulin signaling pathway and cause IR by reducing phosphorylation of insulin receptors and IRS-1. SFAs may also exert lipotoxic effects by increasing inflammation, oxidative stress and endoplasmic reticulum stress, leading to pancreatic β-cell apoptosis and dysfunction.⁷¹ However, most studies about circulating and urine levels of fatty acids in gestational diabetes lack consistent results, and different fatty acid subtypes play different roles in GDM risk and require further validation.

Carnitine

Carnitine, which is synthesized from lysine and methionine, is combined with long-chain acyl-CoA to form acylcarnitine. This process aids delivery of long-chain FAs across the inner mitochondrial membrane during β-oxidation of FAs.⁷² The effects of several acylcarnitine species on GDM vary depending on chain length, and medium-chain acylcarnitine may induce pancreatic β-cell dysfunction.^73,74 By regulating phosphorylation of mTOR, acylcarnitine dysfunction can interfere with insulin signaling and result in IR.³⁸ The urine metabolic fingerprint of GDM in Dudzik's study demonstrated not only changes in amino acids but also abnormalities in carnitine metabolism. The level of carnitine in the urine of GDM patients at 22∼28 weeks of gestation is apparently decreased.²¹ Similarly, low plasma carnitine concentration has been significantly associated with abnormal glucose and lipid metabolism in GDM patients during the second trimester of gestation. Based on prepregnancy BMI, weight gain, and three plasma carnitine metabolites, a prediction model for macrosomia in GDM offspring was established, with an AUC value of 0.88.⁷⁵ Carnitine supplementation boosts the ability of resting and exercising muscles to form acylcarnitine, which contributes to the beneficial effects on metabolic flexibility in individuals with impaired glucose tolerance.⁷⁶ Carnitine is expected to become a potential target for future research in diagnosis and treatment of GDM.

Glycerophospholipid

Glycerophospholipids play a role in biomembrane formation and cell signal transduction and can be further subdivided into phosphatidylcholine (PC), phosphatidylethanolamine, phosphatidylserine, phosphatidylinositol, phosphatidic acid, and cardiolipin, among others.⁷⁷ Many studies have reported glycerophospholipid perturbation in disease states, including prediabetes, T2DM and GDM.78, 79, 80 Sakurai et al. carried out a case-control study in a cohort investigating predictive metabolic biomarkers for GDM in a Japanese population. Urine metabolites were analyzed by hydrophilic interaction chromatography tandem mass spectrometry (HILIC-MS/MS). In urine samples obtained during the first or early second trimester of gestation, ethanolamine, one of the metabolites in glycerophospholipid metabolism, showed an AUC value > 0.8 in the prediction model for discriminating GDM patients from others.²⁵ The glycerophospholipid molecules PC (32:1) and choline ether phospholipid (PCae) (40:4) in serum correlate significantly with 1-h postload glucose (1 h-PG) in the 75 g oral glucose tolerance test (OGTT) and the occurrence of GDM, which also supports involvement of glycerophospholipid metabolism in the pathogenesis of GDM. Together with triglycerides (51:1) and triglycerides (48:1), these four lipid biomarkers raised the AUC value of a predictive model based on maternal age and BMI from 0.69 to 0.74.⁸¹ However, the effect of ethanolamine itself on GDM is not clear and needs further study.

Sphingolipid

Sphingolipids are involved in formation of membrane domains, exosomes and endosomes and modulate cell-cell interactions and cell recognition, such as cell adhesion and migration, cell senescence and death.⁸² Sphingolipids affect pregnancy status, with implications for preeclampsia, intrahepatic cholestasis of pregnancy, GDM and intrauterine growth restriction.⁸³ By adopting the UPLC-MS method for untargeted metabolomics analysis, Yamilé López-Hernández et al. discovered fourteen metabolites that increased dramatically in the urine from GDM subjects during the third trimester, including ceramide (Cer) (d18:0/23:0) and sphingomyelin (SM) (d18:0/22:0).²² Ceramide, which belongs to the sphingolipid family, has been implicated in IR, T2DM and GDM.⁸⁴ Serum Cer (d18:1/24:0) in early pregnancy was higher among women with GDM compared to women without GDM and identified as an independent predictor of GDM.⁸⁵ Increased de novo ceramide synthesis in the endoplasmic reticulum can activate protein phosphatase 2A (PPA2), inactivate protein kinase B (Akt) in the insulin signaling pathway and ultimately inhibit insulin sensitivity and β-cell function.⁸⁶ As one of the most prominent phospholipid components in animal cell membranes, sphingomyelin has been observed to correlate inversely with GDM in several studies.⁶⁸ Furthermore, downregulated sphingolipid metabolism may be an early predictor of the transition from GDM to T2DM. In vivo and in vitro experiments have confirmed that insulin secretion stimulated by glucose is damaged by inhibition of SM metabolism.⁸⁷ The chain length and degree of desaturation of the FA moieties in lipid molecules increase the complexity of their biological roles. In addition, lipids that are endogenously synthesized or obtained from the diet influence their accumulation and/or metabolism and subsequent biological roles.³⁷ Therefore, it is not surprising that there are conflicting views on the pathogenicity and mechanisms of glycerophospholipids and sphingolipids in the development of GDM. New high-resolution metabolomics technology offers broad prospects for solving this problem.

Steroid

Steroid hormones, including estrogen, progestin, androgen and corticosteroid, form a metabolic network that exerts powerful biological activity within the body through specific receptors.⁸⁸ Abnormal steroid hormone metabolism can contribute to various gynecology and obstetrics diseases, such as preeclampsia, polycystic ovary syndrome (PCOS) and GDM.^89,90 Steroid hormones regulate pancreatic function and induce IR, mainly including diabetogenic progesterone, antidiabetic estradiol, androgens related to IR and hyperinsulinemia, neuroactive steroids that affect pituitary functioning, and cortisol, which inhibits the proinflammatory effects of corticotropin-releasing hormone (CRH).^91,92 In the study of López-Hernández, seven steroid derivatives were significantly elevated in GDM patient urine. The dysregulation of the metabolic pathway in steroid hormone biosynthesis that was identified in the pathway topology analysis and pathway enrichment analysis had a close relationship with development of GDM.²² The total metabolic pathway in serum from cholesterol to downstream steroid hormones was increased in GDM condition, especially the estrogen metabolites via 16-pathway. 16α-hydroxyestrone achieved the highest diagnostic performance with an AUC value of 0.85.⁸⁸ Steroid metabolites have the potential to improve GDM diagnosis and further in vitro and in vivo studies on the role of steroid metabolites in pathological process of GDM should be conducted.

Purine metabolites

Purines are among the biomolecules essential for sustaining life and are involved in DNA/RNA synthesis, energy production and signaling transmission. Dysregulation and malfunction of purine and purinergic signaling are involved in the pathophysiology of gout, diabetes mellitus, nervous system diseases, cancers and other diseases.⁹³ Purines and their derivatives in urine are closely related to the etiology of GDM.

Uric acid

Hypoxanthine is a degradation product of ATP and can be converted to xanthine and uric acid by xanthine oxidase (XO). In the aforementioned Law's study, tracking of urine metabolites in the three trimesters of pregnancy also revealed a significant change in purine metabolism in GDM patients. Urine hypoxanthine, xanthine, xanthosine and 1-methylhypoxanthine increased in GDM subjects. As an end product of purine catabolism, uric acid could distinguish GDM patients from healthy pregnant women. Receiver operating characteristic (ROC) analysis of key purine metabolites showed that xanthine had a particularly good capacity to diagnose GDM, with an AUC value > 0.8, in each trimester.²⁰ In some studies on plasma purine degradation metabolites, similar results have shown that xanthine, hypoxanthine, and inosine monophosphate (IMP) are elevated in GDM patients.^94,95 A meta-analysis including 11 observational studies found that the level of serum uric acid in GDM patients was higher and correlated positively with an increased risk of GDM, especially in the first trimester.⁹⁶ Therefore, a model to elucidate the role of tryptophan and purine metabolism in the pathogenesis and pathophysiology of GDM is hypothesized. Imbalance of proinflammatory and anti-inflammatory cytokines induces chronic low-grade inflammation in GDM and upregulates the Trp-KYN pathway by activating IDO1. Increased synthesis of xanthurenic acid in the Trp-KYN pathway, which has diabetogenic effects, leads to pancreatic β-cell dysfunction and hyperglycemia. Furthermore, hyperglycemia accelerates purine nucleotide synthesis, thereby stimulating nucleotide breakdown and increasing the concentration of nucleotide degradation products, including superoxide molecules and uric acid. Reactive oxygen species (ROS) and uric acid lead to further deterioration of the condition. Excessive uric acid induces oxidative stress, ROS accumulation and mitochondrial injury in hepatocytes. Moreover, a high concentration of uric acid also contributes to local inflammation in adipocytes, with a reduction in adiponectin and induces IR.⁴⁷

8-Hydroxy-2-deoxyguanosine

Oxidative stress is a state of imbalance between overproduction of ROS and inadequate antioxidant defense mechanisms. Accumulation of ROS, including superoxide anion radicals, hydrogen peroxide and hydroxyl radicals, leads to oxidative damage to proteins, lipids, and DNA. 8-Hydroxy-2-deoxyguanosine (8-OHdG) is generated by interaction of hydroxyl radicals with guanine on DNA strands and is stably excreted in the urine; it is one of the most extensively studied biomarkers of oxidative damage of DNA.⁹⁷ Several diseases have been reported to correlate with high urine concentrations of 8-OHdG, such as diabetes, hypertension, CVD and chronic obstructive pulmonary disease (COPD). Aging, smoking, or occupational exposure to physical, chemical or biological substances may also increase the level of 8-OHdG.⁹⁸ Qiu et al. conducted a case-control study comparing the urine 8-OHdG level of GDM and non-GDM pregnant women at 16 weeks of pregnancy. The risk of developing GDM in patients with a urine 8-OHdG concentration >8.01 ng/mg creatinine was 3.79-fold higher than that in patients with a urine 8-OHdG concentration <4.23 ng/mg creatinine.²⁶ This result supports that oxidative stress is a contributor to GDM and needs to be kept under strict control.⁹⁹ Furthermore, some studies have found that iron overload is associated with an increased risk of GDM, with possible mechanisms including increased oxidative stress, which contribute to IR and inadequate insulin secretion.¹⁰⁰ In addition to collecting urine to assess changes in 8-OHdG, Tuna et al. collected blood to measure indicators related to blood glucose and iron status at 11∼14 weeks of gestation and 24∼28 weeks of gestation, including fasting blood glucose (FBG), HbA1c, ferritin and hemoglobin levels. Compared with the control group, urine 8-OHdG levels were higher in the GDM group in the first trimester. After stratification by serum ferritin levels in the first trimester, women in the ≥50th percentile (≥13 ng/mL) had higher 8-OHdG levels than women in the <50th percentile. Notably, a significant positive association between 8-OHdG and ferritin and 1 h-PG in the OGTT was observed in the second trimester.²⁷ Overall, interaction between hyperglycemia, iron status and DNA oxidative damage may play a role in the pathogenesis of GDM and is a direction that needs to be highlighted in future research.

Urine proteomics

The urine proteome originates from the urogenital tract and glomerular plasma filtration and thus contains potential biomarkers associated with general health and specific urogenital diseases.¹⁰¹ A proteomic biomarker is defined as a specific peptide or protein related to a specific state, for example, initiation, feature or progression of a disease or response to treatment.¹⁰² In several diseases, abnormal proteins have been found in urine, mainly focusing on urogenital system disorders.^103,104 Urine biomarkers subtype 22 of the family of nuclear matrix proteins and human complement factor H-related protein have been approved by FDA for diagnosis and surveillance of bladder cancer.¹⁰⁵ MS-based urine proteomics facilitates biomarker discovery. Differences in protein or peptide concentrations between a control group and disease group are compared by means of label-free or label-based relative quantification methods and lead to selection of candidate biomarkers. Confirmation of these biomarkers by targeted absolute quantitative analysis is needed prior to clinical validation and application.¹⁰² In the biomarker discovery phase, two-dimensional electrophoresis (2DE), LC and CE coupled to MS platforms are common technologies; in the validation phase, both traditional immunological assays such as ELISA and novel techniques such as selected/multiple reaction monitoring (SRM/MRM) technology are frequently used.¹⁰⁶ ELISA has high sensitivity and specificity, but is also challenged by antibody availability and the cost of assay development for new target analytes. MS-based absolute quantification has emerged as a successful alternative to immunoassays, offering lower cost, shorter turnaround time, and greatly improved throughput.¹⁰⁷ With the application of urine proteomics, a number of potential protein biomarkers for GDM have been identified in recent years.

Liver-type fatty acid-binding protein

A member of the fatty acid-binding protein superfamily, liver-type fatty acid-binding protein (L-FABP) is synthesized in the proximal renal tubules and is involved in FA metabolism by binding and transporting FAs to mitochondria and peroxisomes, where they are oxidized to produce energy.¹⁰⁸ Elevated levels of L-FABP in urine can identify diabetic nephropathy in patients with T1DM and T2DM and can also predict progression and severity of diabetic nephropathy.¹⁰⁹ Because GDM and diabetes have similar pathogeneses, kidney damage, a common complication of T2DM, also occurs in GDM. Some studies have shown that women with GDM have an increased risk of kidney disease after delivery, including CKD and end-stage renal disease.^110,111 Urine L-FABP measured by the ELISA kit was higher in pregnant women with GDM than in healthy pregnant women and healthy nonpregnant women and correlated positively with HbA1c and FBG.²⁸ Urine L-FABP may become an indicator of renal tubular injury under the stress of hyperglycemia during pregnancy.

Inter-alpha-trypsin inhibitor heavy chain H4

Inter-alpha-trypsin inhibitor heavy chain H4 (ITIH4) is a glycoprotein synthesized in the liver and belongs to the inter-alpha-inhibitor (ITI) family.¹¹² It has been implicated in the pathogenesis of several diseases, such as various cancers, preeclampsia and recurrent pregnancy loss.112, 113, 114, 115, 116, 117, 118 Hu et al. identified two urine peptides that exhibited noteworthy differences between the healthy pregnancy group and GDM group through utilization of matrix-assisted laser desorption ionization time-of-flight (MALDI-TOF) MS.²⁹ One of the two urine peptides was ITIH4, which was detected at higher levels in GDM patients. Tracking of the dynamic changes in urine ITIH4 levels demonstrated no significant difference between the second and third trimesters, providing a basis to study the alterations in secretion patterns of urine peptides during GDM. In the GDM group, the subgroup with high fasting plasma glucose (FPG) had higher urine ITIH4 levels than the subgroup with low FPG. Secretion of ITIH4, an acute-phase inflammation-associated protein, is increased with GDM exacerbation, vascular inflammation, and immune dysfunction. Moreover, the trend of urine ITIH4 levels in GDM patients is in parallel with the reciprocal of neonatal fasting plasma glucose (NFPG) and neonatal weight (NW), indicating that it might be a predictor not only of the severity of glucose metabolism dysfunction in patients with GDM but also of neonatal hypoglycemia and macrosomia.

Coagulation factor IX

Coagulation factor IX is a vitamin K-dependent glycoprotein that mediates the coagulation response by receiving both intrinsic and extrinsic activation signals.¹¹⁹ In Hu's study, peptides in urine collected during the second and third trimesters were analyzed by MALDI-TOF MS. Coagulation factor IX, TBC1 family member 5 isoform a (TBC1D5a) and immunoglobulin kappa constant were higher in GDM patients than in nonpregnant women and healthy pregnant women. A model containing FPG, HbA1c and three urine peptides can diagnose GDM, with an AUC value of 0.785.³⁰ Pregnancy is often in a physiologic hypercoagulable state. However, under the burden of hyperglycemia, it becomes a pathological coagulation/fibrinolysis disorder.¹²⁰ Proteomics analysis of serum, blood exosomes and placental villi also led to similar results, namely, that the imbalance of coagulation and fibrinolysis is involved in the pathophysiology of GDM.121, 122, 123 Hyperglycemia-induced endothelial damage upregulates the coagulation system in pregnant women with GDM, and GDM patients have a higher risk of arteriovenous thrombosis.¹²⁴ In the future, urine peptide detection kits based on urine coagulation-related peptides may be employed for diagnosis of GDM.

Conclusion

Metabolomics and proteomics related to GDM study is in a period of vigorous development. Urine amino acid metabolites, lipid metabolites, purine metabolites, key proteins and peptides show good potential in predicting and diagnosing GDM and are associated with the etiology of GDM (Fig. 1). Further large-scale and standardized longitudinal research is necessary to validate and extend current findings. Urine biomarkers of GDM will be a powerful tool in improving maternal and fetal health and interrupting the intergenerational transmission of metabolic diseases.

Fig. 1.

Potential urine biomarkers for prediction and diagnosis of GDM by using metabolomics and proteomics.

Outstanding questions

Although numerous studies have identified changes in urine components in pregnant women with GDM that are expected to serve as predictive or diagnostic biomarkers, there are several questions that need to be addressed before translating these findings from the laboratory to clinical application.

Metabolomics and proteomics techniques have been employed in studies about urine biomarkers of GDM, yielding a substantial quantity of data. However, data repeatability remains low, with certain outcomes appearing to be contradictory. Disparities among such findings may arise from variations in GDM diagnostic criteria, ethnicity, population size, urine sample preparation and analysis platforms. Most of the studies on urine biomarkers of GDM that have been conducted are small in scale and only detected urine in a certain period during pregnancy, without whole-pregnancy tracking. In general, the composition of urine is susceptible to interference from various factors, such as diet, exercise, and medication, which may hinder the interpretation of research results.

Therefore, to validate these identified urine biomarkers of GDM and discover new biomarkers, large-scale, multicenter, multiethnic, dynamic surveillance and prospective cohort studies are necessary in the future. Although spot or morning urine collection is a convenient approach to sample collection, 24-h urine collection is recommended because it provides a more precise and accurate measurement of urine components. Combining changes in GDM biomarkers in urine with their corresponding changes in blood or feces provides insight into their metabolic trajectories in the body and well reflects their potential for predicting or diagnosing GDM. Participant recruitment, sample preparation and storage procedures, biomarker measurement and analysis, and elaboration and evaluation of results all require standardized designs to collect meaningful and comparable data. Rapid advancement in analytical technology and the upgrade of metabolic databases, bioinformatics and artificial intelligence have also provided assistance in further searching for urine biomarkers of GDM. By using sophisticated biology computational tools to integrate urine metabolomics, proteomics, genomics and transcriptomics data, such powerful integrated analysis will elucidate the pathogenesis and pathophysiology of GDM and provide targets for precise therapy.

Contributors

Jie Yu, Xinhua Xiao and Qian Zhang conceived the paper. Jie Yu, Jing Ren and Yaolin Ren wrote the original draft. Jie Yu, Yifan Wu and Yuan Zeng polished the manuscript and prepared tables and figures. Xinhua Xiao and Qian Zhang contributed to the final revision and editing of this manuscript. All authors have read and approved the published version of manuscript.

Declaration of interests

The authors declare no conflicts of interest.