Daniela Duarte, Maria do Céu Almeida, Pedro Domingues, and Ana M. Gil
Main prenatal disorders and challenges related to clinical diagnosis
Pregnancy-related disorders comprise maternal and fetal disorders, each condition posing different challenges in terms of diagnosis and clinical management. Table 20.1 lists the main disorders that arise during pregnancy and their characteristics. A more comprehensive and clinical description of such disorders is out of the scope of this text and may be found elsewhere (1,2). Much of the research on improving clinical diagnosis of prenatal disorders has searched for early, predictive, and effective (high sensitivity and specificity) biomarkers of the disorder and of related subsequent health complications (Table 20.1). The interest in less invasive methods of diagnosis and followup has led to increasing exploitation of biofluids (mainly blood), with a growing interest in more non-invasive biological matrixes such as urine and saliva.
Table 20.1 Pregnancy-related complications: Definitions, onset/diagnosis time, incidence, and short- and long-term complications
In particular relating to omic sciences in prenatal health research, many disorders have been studied at their prediagnostic stages, usually in small cross-sectional cohorts. Human biofluids pose the challenge of interindividual variability; hence, efforts to study larger longitudinal cohorts are emerging so that potential biomarkers are based on intraindividual response. In addition, it is important to assess potential biomarkers across distinct cohorts, to account for local phenotypic effects. An added challenge relates to the use of different omic sciences usually in separate studies, leaving an open area of interest ahead related to omics correlation.
Overall applications of metabolomics and proteomics in prenatal health
Metabolomics consists of the global analysis of metabolites in biofluids, tissues, and cells (17), to study biological response to a selected perturbation, e.g., disease. Usually metabolomics involves the use of nuclear magnetic resonance (NMR) spectroscopy or tandem mass spectrometry (MS) methods (Figure 20.1). In this context, NMR- and MS-based methods are complementary (18) since the former suits more holistic and untargeted studies and the latter, mostly due to its higher sensitivity (<pmolar, compared to mmolar in NMR), is usually chosen for studies targeting specific compound families. The amount of complex data thus generated requires multivariate analysis (MVA), either unsupervised, e.g., principal component analysis (PCA), or supervised, e.g., partial least square discriminant analysis (PLS-DA), in tandem with univariate statistical analysis, necessarily requiring statistical validation procedures (19). Metabolomics generates putative hypotheses of deviant metabolic pathways, which also require biological validation at a later stage of the metabolomics strategy. Figure 20.2 shows the number of metabolomics publications focused on prenatal research, per year and condition (Figure 20.2a), as well as per sample type (Figure 20.2b). It becomes apparent that gestational diabetes mellitus (GDM, carbohydrate intolerance with onset in pregnancy) and preterm birth (PTB, birth before 37 gestational weeks, g.w.) have been the most extensively studied conditions, with maternal blood addressed in approximately 40% of all papers, with urine following in just under 20% of reports (Figure 20.2b).
Figure 20.1 General workflows employed in metabolomic and proteomic studies. (GC, gas chromatography; LC, liquid chromatography; MS, mass spectrometry; NMR, nuclear magnetic resonance; 1D/2D PAGE, unidimensional/bidimensional polyacrylamide gel electrophoresis.)
Proteomics is the study of the entire set of proteins expressed by a cell, tissue, or whole organism, measured at a given time and under specific conditions (20). Proteome analysis usually comprises an initial step to isolate the protein mixture from the remaining sample components. This is followed by protein digestion, usually with trypsin, sometimes preceded by gel separation. The digested sample is subjected to online high-performance liquid chromatography-mass spectrometry (HPLC-MS) peptide separation, using MS and MS-MS for quantification and identification, respectively (Figure 20.1). Bioinformatics tools are also required to handle the resulting data, leading on to hypotheses generation and subsequent validation. In prenatal health, proteomic studies have been less extensively employed compared to metabolomics, having particularly addressed preeclampsia (PE) and chromosomal disorders (CD) (Figure 20.3a). Similarly to metabolomics, maternal blood has been the preferred biological matrix (close to 50% of all papers), whereas urine has received little attention so far (Figure 20.3b).
Figure 20.2 Metabolomics in prenatal health research. (a) Number of publications per year and condition. (b) Number and percentage of papers per sample type. Information obtained through Web of Science ( www.webofknowledge.com), PubMed (www.ncbi.nlm.nih.gov/pubmed/) and articles cited in selected papers (* updated to January 2019). (CD, chromosomal disorders; CVF, cervicovaginal fluid; FG, fetal growth; FM, fetal malformations; GDM, gestational diabetes mellitus; NB, newborn blood; PE, preeclampsia; PROM, premature rupture of membranes; PTB, preterm birth.) Others conditions comprise intrahepatic cholestasis of pregnancy, bronchopulmonary dysplasia, miscarriage, maternal obesity, intra-amniotic infection, human cytomegalovirus, jaundice, prenatal exposure to pollutants. Other sample types include bronchoalveolar lavage fluid and meconium.
Figure 20.3 Proteomics in prenatal health research. (a) Graphs of the number of proteomics papers comprising each condition as a function of the publication year. (b) Number, alongside with percentage of papers exploring each sample type. This information was obtained through consultation of the Web of Science (www.webofknowledge.com), Pubmed (www.ncbi.nlm.nih.gov/pubmed/) and articles cited in selected papers. It is updated until January 2019 (*). AF, amniotic fluid; CVF, cervicovaginal fluid; CSF, cerebrospinal fluid; MB, maternal blood; MS, maternal saliva; MU, maternal urine; NB, newborn blood; NS, newborn saliva; NU, newborn urine; PTB, preterm birth; UCB, umbilical cord blood. Others conditions comprise intra-amniotic infection, human cytomegalovirus, respiratory distress, multiple sclerosis and prenatal exposure to pollutants. Other sample types include amniocytes, adipose tissue, skeletal muscle, colostrum and fetal heart tissue.
Urine metabolomics in prenatal health
Noninvasive biofluids such as urine and saliva have increasingly been the focus of omic sciences in disease research. In prenatal research, analysis of maternal urine has been reported in approximately 30 studies (Figure 20.2b), most of which related to the main prenatal disorders (as listed in Table 20.1), and the remaining addressing other issues explored to lesser extents, namely, spontaneous miscarriage, intrahepatic cholestasis of pregnancy, maternal obesity, and pollutant effects. Table 20.2 summarizes the main characteristics of the main studies, which cover a wide range of cohort characteristics and analytical strategies, leading to a considerable amount of knowledge on metabolome variations for each disorder. NMR and MS have been used to comparable extents, usually in separate studies, and Figure 20.4 shows examples of typical records of urine NMR and ultra-performance liquid chromatography (UPLC) leading up to MS detection.
Figure 20.4 Example of data in urine metabolomics. (a) 1H-NMR spectrum of second trimester urine of a healthy pregnant woman. 1, β-hydroxybutyrate; 2, 3-aminoisobutyrate; 3, lactate; 4, threonine; 5, alanine; 6, γ-aminobutyrate; 7, succinate; 8, citrate; 9, dimethylamine; 10, creatine; 11, creatinine; 12, trimethylamine N-oxide; 13, betaine; 14, glycine; 15, guadinoacetate; 16, trigonelline; 17, glucose; 18, histidine; 19, phenylacetylglycine; 20, hippurate; 21, formate; 22, N-methyl-nicotinamide. (b) Positive ion total ion chromatographs from a healthy pregnant woman in different pregnancy trimesters. 1T, 2T, 3T: first, second and third trimesters. (Source: [a] Ref. 21; [b] Ref. 31.)
Gestational diabetes mellitus
Most GDM metabolomic studies through urine have aimed at finding predictive metabolic biomarkers before diagnosis, adding to reports at the time of GDM diagnostics and including two studies across pregnancy, one of these of a longitudinal nature (31) (Table 20.2). In an initial study (21), a prediagnostic GDM group was compared to those described by other conditions (preterm premature rupture of membranes [PPROM], PTB, fetal malformations [FM], and CD). Upregulation of 3-hydroxyisovalerate (3-HIVA), and 2-hydroxyisobutyrate (2-HIBA) were found to be apparently specific of GDM before diagnosis, possibly reflecting decreased biotin status. Variations in common to all conditions (probably as a result of stress effects) comprised increased levels of choline, N-methyl-2-pyridone-5-carboxamide (2-Py), and N-methyl-nicotinamide (NMND). A later similar study addressed a larger cohort (25) and revealed prediagnostic increased levels of choline, glucose (possible glycemia), NMND, and xylose, and decreases in 4-hydroxyphenylacetate (4-HPA) and hippurate, reflecting the importance of gut microflora in GDM development (25). The somewhat different observations compared to the previous study (apart from the common choline increase) illustrate the importance of cohort size in determining reliable metabolomic results. This was confirmed in a study of a multiethnic cohort, which failed to confirm any metabolic biomarker for prediagnostic GDM apart from an increasing tendency in citrate levels with increasing hyperglycemia severity (22). Untargeted MS-based studies enabled increased choline, ethylmalonate, and pyruvate levels to be detected in prediagnostic GDM, along with decreased adipate levels (23,26). A more recent MS longitudinal study (31) identified several more metabolic alterations during the whole pregnancy period (Table 20.2) and, through pathway analysis, suggested tryptophan metabolism as the most affected pathway in GDM (Figure 20.5). Furthermore, downregulated ethanolamine and upregulated 1,3-diphosphoglycerate were found as potentially predictive of GDM in a Japanese cohort (33).
Figure 20.5 Results of metabolomics in GDM research. Output of pathway analysis of metabolites selected by multivariate analysis and differentiating GDM and healthy controls.
(Source: Ref. 31.)
In relation to GDM diagnosis, five urinary metabolomic studies were performed (27–30,32). Increased levels of several amino acids were observed in general, together with decreased carnitine. In the most recent studies (30,32), a large number of changes were observed in the levels of several hormones and related compounds (Table 20.2); however, some of these changes lost significance upon correction for multiple comparisons (32).
Additionally, the metabolic response to GDM treatment (either diet alone or insulin) was evaluated by NMR (29). In this study, different sets of varying metabolites were suggested to correlate either to responsive or resistant metabolic pathways, as well as evidenced possible metabolic side effects of treatment. The knowledge of excretory metabolome response to treatment unveils the possibility of monitoring therapy in personalized protocols.
To our knowledge, all PTB metabolomic studies have searched for urinary predictive biomarkers (Table 20.2), mostly considering small cohorts. Amino acid and choline metabolism disturbances have been suggested, with basis on increased levels of 2-HIBA, 3-methylhistidine, and choline and decreased 4-HPA levels (the latter observed in common with other prenatal conditions), as viewed by NMR (21,25). However, an MS study of a similar cohort did not find any significant metabolite changes (23). A more recent report addressed maternal urine of a larger pre-PTB cohort, split into pre-medically induced (pre–iPTB) and pre-spontaneous (pre–sPTB), during the first trimester (34). Increased levels of lysine, a steroid conjugate, 2-Py, and decreases in TMAO, glycine, and formate were suggested to help predict sPTB. In relation to iPTB, decreased levels of phenylacetylglutamine and increased levels of N-acetyl glycoproteins were found in first trimester maternal urine. The latter was suggested to be linked to inflammation preceding iPTB.
To the best of our knowledge, only NMR metabolomics has been employed to search for urinary predictive biomarkers of PE in early pregnancy (25,42) and at the time of diagnosis (35) (Table 20.2). Pre-PE metabolic disturbances were initially suggested to affect mostly energy metabolism (25). A second pre-PE study, in the first trimester, compared pre–PE cases with subjects affected by gestational hypertension (pre-GH) (42), thus addressing the important issue of specificity. PE and GH cases were predicted with 51.3% and 40% sensitivity, respectively, and the most important varying metabolites in women who later developed either condition were hippurate (decreased) and creatinine (increased). The urinary metabolic profile associated to pre-PE also included increased levels of glycine and 4-DTA, and decreased lactate and creatine. Receiver operating characteristic (ROC) curve analysis (a measure of sensitivity and specificity) provided higher accuracy (80.7%) when combining hippurate/creatinine levels with maternal age, mean-arterial pressure (MAP), and uterine artery pulsatility index (UtAPI) (Figure 20.6).
Figure 20.6 ROC curve comparison from logistic regression analyses. Prediction of preeclampsia using logistic regression, with risk of preeclampsia as dependent variable and several maternal characteristics as independent variables. (MAP, mean arterial pressure; UtAPI, uterine artery pulsatility index.) (Source: Ref. 42, under Creative Commons.)