Metabolic Biomarkers of Prenatal Disorders: An Exploratory NMR

Jun 8, 2011 - Showing 1/6: pr200352m_si_001.pdf. figshare. 1 / 6. Share ..... American Journal of Obstetrics and Gynecology 2015, 213 (6) , 841.e1-841...
0 downloads 0 Views 2MB Size
ARTICLE pubs.acs.org/jpr

Metabolic Biomarkers of Prenatal Disorders: An Exploratory NMR Metabonomics Study of Second Trimester Maternal Urine and Blood Plasma Sílvia O. Diaz,†,‡ Joana Pinto,†,‡ Gonc) alo Grac) a,‡ Iola F. Duarte,‡ Antonio S. Barros,§ Eulalia Galhano,^ Cristina Pita,^ Maria do Ceu Almeida,^ Brian J. Goodfellow,‡ Isabel M. Carreira,|| and Ana M. Gil*,‡ ‡

CICECO Department of Chemistry, Campus Universitario de Santiago, University of Aveiro, 3810-193 Aveiro, Portugal QOPNA Department of Chemistry, Campus Universitario de Santiago, University of Aveiro, 3810-193 Aveiro, Portugal Cytogenetics and Genomics Laboratory, Faculty of Medicine, University of Coimbra, Portugal and CENCIFOR - Forensic Science Centre, Portugal ^ Maternidade Bissaya Barreto, Centro Hospitalar de Coimbra, 3000 Coimbra, Portugal

J. Proteome Res. 2011.10:3732-3742. Downloaded from pubs.acs.org by IOWA STATE UNIV on 01/28/19. For personal use only.

)

§

bS Supporting Information ABSTRACT: This work describes an exploratory NMR metabonomic study of second trimester maternal urine and plasma, in an attempt to characterize the metabolic changes underlying prenatal disorders and identify possible early biomarkers. Fetal malformations have the strongest metabolic impact in both biofluids, suggesting effects due to hypoxia (leading to hypoxanthine increased excretion) and a need for enhanced gluconeogenesis, with higher ketone bodies (acetone and 3-hydroxybutyric acid) production and TCA cycle demand (suggested by glucogenic amino acids and cis-aconitate overproduction). Choline and nucleotide metabolisms also seem affected and a distinct plasma lipids profile is observed for mothers with fetuses affected by central nervous system malformations. Urine from women who subsequently develop gestational diabetes mellitus exhibits higher 3-hydroxyisovalerate and 2-hydroxyisobutyrate levels, probably due to altered biotin status and amino acid and/or gut metabolisms (the latter possibly related to higher BMI values). Other urinary changes suggest choline and nucleotide metabolic alterations, whereas lower plasma betaine and TMAO levels are found. Chromosomal disorders and pre-preterm delivery groups show urinary changes in choline and, in the latter case, in 2-hydroxyisobutyrate. These results show that NMR metabonomics of maternal biofluids enables the noninvasive detection of metabolic changes associated to prenatal disorders, thus unveiling potential disorder biomarkers. KEYWORDS: maternal urine and plasma, prenatal disorders, fetal malformations, gestational diabetes, preterm, NMR spectroscopy, metabonomics

’ INTRODUCTION Prenatal diagnostic methods include a variety of both invasive and noninvasive procedures, with emphasis on routine ultrasound exams, which enable the monitoring of fetal morphological development and, in combination with maternal serum markers (PAPP-A and hCGβ), the detection of chromosomal disorders such as trisomy 21, as early as 11 13 weeks.1 For higher risk pregnancies, invasive techniques such as chorionic villus sampling (CVS), amniocentesis, and cordocentesis (or fetal blood sampling) are employed to determine the karyotype for the diagnosis of chromosomal anomalies or specific genetic diseases.2 However, despite their importance in prenatal diagnostics, some of these procedures still carry some risk of miscarriage, infection, or premature birth.2 Furthermore, other disorders such as gestational diabetes mellitus (GDM), preeclampsia (PE), preterm delivery (PTD), Small for Gestational Age (SGA), or IntraUterine Growth Restriction (IUGR), for r 2011 American Chemical Society

which often no adequate predictive methods are available, may pose a significant risk to mother and fetus later in pregnancy. Metabonomics is a possible holistic avenue in the quest for new and earlier biomarkers of prenatal health. This is a strategy strongly based on Nuclear Magnetic Resonance (NMR) spectroscopy and/or Mass Spectrometry (MS), which can follow the metabolic responses of living systems to disease, toxicological or nutritional stimuli by studying metabolic profiles of biofluids and other biological samples.3 Indeed, the potential of biofluids metabolic profiling for prenatal biomarker identification is being increasingly recognized. In this context, NMR of amniotic fluid has addressed studies of fetal maturity,4,5 effects of gestational age6,7 and of several disorders, for example, PE, fetal malformations (FM), GDM, PTD, and premature rupture of membranes Received: April 15, 2011 Published: June 08, 2011 3732

dx.doi.org/10.1021/pr200352m | J. Proteome Res. 2011, 10, 3732–3742

Journal of Proteome Research

ARTICLE

Table 1. List of Urine and Blood Plasma Samples with Corresponding Sample Numbers and Maternal and Gestational Age Ranges, at the Time of Sample Collection maternal ages

gestational weeks

no. urine samplesa

Control

28 41

16 21

25 (23.9; 14)

20 (22.7; 8)

Fetal Malformations (FM) Prediagnostic GDM

26 40 30 44

14 25 16 22

26 (24.4; 13) 29 (26.5; 21)

27b (25.3; 13) 14 (26.6; 9)

Chromosomal disorders

25 47

14 25

23 (27.3; 15)

23 (26.1; 16)

Pre-preterm delivery (pre-PTD)

26 41

16 21

17 (26.7; 8)

4c

Pre-premature rupture of the membranes (pre-PROM)

27 42

15 21

38 (23.6; 20)

18 (22.7; 9)

sample group

no. blood plasma samplesa

a

Numbers in brackets indicate average BMI and number of samples for which BMI information is available. b This group included 3 samples of the 3rd trimester: two corresponding to central nervous system malformations (32 and 35 gestational weeks) and one to a cardiac malformation (32 gestational weeks). c Insufficient number of samples for study.

(PROM).4,8 10 In addition, several MS reports on amniotic fluid have emerged recently, comprising an assessment study of the risk of PTD11 and targeted analysis for gestational age effects12 and cholesterol biosynthesis defects.13 In spite of the great potential of prenatal urine and blood for the development of novel less-invasive diagnosis and monitoring methods, the exploitation of metabolite profiling of these biofluids is less developed. Targeted metabolite evaluation of plasma has sought correlations to GDM, either focusing on surrogate metabolite markers of oxidative stress in tandem with expected biomarker proteins14 or by examining the changes in plasma lipids and lowdensity lipoproteins.15 In addition, a large-scale study has shown that second trimester serum biomarkers such as alpha-fetoprotein, human chorionic gonadotropin and unconjugated estriol may have predictive value for preterm birth.16 Most metabonomics studies on prenatal blood have, however, focused on PE, comprising the NMR study of the blood plasma of 11 PE women and 11 controls, leading to the observations of low lipid and ketone body levels17 and different aromatic amino acid profiles18 for PE women. Also, a larger scale study of 15 gestational week’s blood plasma (i.e., at prediagnostic stages) by ultra high performance liquid chromatography and mass spectrometry (UPLC MS)19 has identified 14 metabolites with predictive power for later PE. However, to our knowledge, no such untargeted approach has been used for other prenatal disorders, either by NMR or MS metabonomics. The same seems to apply to the analysis of prenatal urine where some work has been carried out in targeted metabolite analysis in relation to GDM,20 steroid sulfatase deficiency (STSD),21 and the effects of oxidative stress on pregnancy outcome.22 This work describes an exploratory NMR-based metabonomics study of second trimester maternal urine and blood plasma, in an attempt to correlate biofluid metabolic changes with suspected/diagnosed prenatal disorders such as FM and chromosomal disorders, as well as with disorders developed later in pregnancy, such GDM, PTD, and PROM (PE was not included due to insufficient number of samples available). Some metabolite variations have been identified as potential disease signatures, thus unveiling the potential of metabolite markers for clinical use as prenatal diagnostic and prognostic tools. Furthermore, comparison of the effects of disease on the two biofluids gives interesting insight into the relative predictive powers of urine and plasma and possible matrix of choice for future improved metabonomics models and targeted analysis. The tandem use of this information with the existing clinical methods is expected to improve pregnancy monitoring and outcome.

’ EXPERIMENTAL SECTION Samples

Urine and blood plasma samples were collected at the time of amniocentesis (14 25 gestational weeks), performed for pregnant women aged >35 or based on medical history. All pregnancies were followed until birth, eventually defining several sample groups according to their clinical characteristics: controls (healthy pregnancies throughout), fetal malformations (FM), prediagnostic gestational diabetes mellitus (GDM) (for women diagnosed with GDM later in their pregnancy), pre-preterm delivery (PTD) (for women who gave birth prior to 37 g. w.), pre-premature rupture of the membranes (PROM) (for women who had prelabor rupture of membranes after 37 g. w.), and chromosomal disorders (generally diagnosed ca. 2 weeks after amniocentesis). For other disorders of interest, for example, preeclampsia, an insufficient number of samples were gathered up to the time of this work. Table 1 summarizes the number of samples in each sample group as well as the corresponding maternal age, gestational age, and body mass index (BMI) at the beginning of pregnancy (unfortunately, the latter information is only known for a limited number of subjects thus hindering a statistically meaningful evaluation of the role of BMI, at this stage). Both FM and chromosomal disorder groups are heterogeneous regarding disorder type, the former comprising malformations of the central nervous system, cardiac, urogenital, soft tissues and pulmonary (also including 3 plasma samples collected at 32 and 35 gestational weeks), and the latter comprising cases with unbalanced chromosome complements (e.g., trisomy 18, 21, deletion and mosaicism) and cases with balanced ones (e.g., carriers of translocations and inversions). All urine and blood samples were collected under nonfasting conditions, due to the medical restrictions in controlling/ limiting pregnant women’s diet up to the time of sample collection. This fact, as well as other potential confounding variables (e.g., lifestyle, environmental, cultural, genetic) may be at the basis of some sporadic abnormal NMR profiles noted for a few control samples. These samples, not representative of the bulk, were removed from the control group. It should be noted that the subjects who donated urine were not necessarily the same who donated blood plasma, although an overlap of 61 subjects (in a total of 198) occurred. This research was carried out under ethical committee approval and informed consents from each subject participating in the study. All clinical and additional metadata (e.g., BMI at beginning of pregnancy, medical history, and lifestyle habits such as smoking and diet) were obtained from obstetrical and neonatal medical records, as 3733

dx.doi.org/10.1021/pr200352m |J. Proteome Res. 2011, 10, 3732–3742

Journal of Proteome Research

ARTICLE

Figure 1. 1H NMR spectra (500 MHz) recorded for 2nd trimester urine and plasma of a healthy pregnant woman: (a) urine, standard 1D spectrum; (b) blood plasma, standard 1D (top), T2-edited (middle) and diffusion-edited (bottom) spectra. Legend: (1) β-hydroxybutyrate, (2) 3-aminoisobutyrate, (3) lactate, (4) threonine, (5) alanine, (6) γ-aminobutyrate (GABA), (7) succinate, (8) citrate, (9) dimethylamine, (10) creatine, (11) creatinine, (12) trimethylamine N-oxide (TMAO), (13) betaine, (14) glycine, (15) guadinoacetate, (16) trigonelline, (17) glucose, (18) histidine, (19) phenylacetylglycine, (20) hippurate, (21) formate, (22) N-methyl-nicotinamide, (23) lipid CH3, (24) lipid (CH2)n, (25), glycoproteins (26) choline, (27) lipid CHdCH, (28) valine, (29) glutamine, (30) tyrosine, (31) lipid CH2CH2CO, (32) lipid CH2CHdCH, (33) lipid CHdCHCH2CHdCH, (34) albumin lysyl ε-CH2.

well as from an individual questionnaire filled in at the time of sample collection. Urine samples (50 mL) were frozen at 20 C for up to 2 h and then stored at 80 C after collection and until NMR analysis. Before analysis, samples were thawed at room temperature and 800 μL were centrifuged (4500 g, 25 C, 5 min). Then, 60 μL of 1.5 M KH2PO4/D2O phosphate buffer pH 7, 0.1%Na+/3-trimethylsilylpropionic acid (TSP) were added to 540 μL of supernatant, followed by buffering to pH 7.00 ( 0.02 with KOD (4 M) or DCl (4 M). The mixture was again centrifuged (4500 g, 25 C, 5 min) and 550 μL were transferred to a 5 mm NMR tube. For blood plasma, whole blood (8 9 mL) was collected into sodium heparin tubes and centrifuged (1500 g, 25 C, 10 min) not more than 30 min after collection. Supernatants were frozen at 20 C for up to 2 h and then stored at 80 C until NMR analysis. Before analysis, samples were thawed at room temperature and 400 μL of saline solution (NaCl 0.9% in 10% D2O) were added to 200 μL of plasma. The mixture was then centrifuged (4500 g, 25 C, 5 min) and transferred into 5 mm NMR tubes. NMR Spectroscopy

NMR spectra were recorded on a Bruker Avance DRX 500 spectrometer equipped with an actively shielded gradient unit with a maximum gradient strength output of 53.5 G/cm, at 300 K. For urine, standard 1D spectra were acquired, using a noesy 1D pulse sequence with mixing time (tm) 100 ms, a fixed 3 μs t1 delay, and water suppression during relaxation delay and mixing time. One-hundred twenty-eight transients were acquired into 64k complex data points, with a 10080.65 Hz ppm spectral width (SW), a 4 s relaxation delay and a 3.25 s acquisition time. For each blood plasma sample, three 1D 1H NMR spectra were obtained: a standard spectrum, a T2-edited spectrum and a diffusion-edited spectrum. Standard 1D spectra were acquired using a noesy1D pulse sequence with tm = 100 ms, a fixed 3 μs t1

delay and water suppression during relaxation delay (4 s) and mixing time. 1D T2-edited spectra were acquired with the RD90-{τ-180-τ}n-acquire pulse sequence, with water presaturation, n = 80, τ = 400 μs and a total spin spin relaxation time (2nτ) of 64 ms. 1D diffusion-edited spectra were recorded using the bipolar pulse longitudinal eddy current delay (BPPLED) pulse sequence, using sine gradients with 2 ms duration and 90% of the maximum gradient strength (48.15 G/cm) and a 100 ms diffusion time. All plasma 1D spectra were acquired with 32k complex data points, 10330.58 Hz spectral width (SW), and 4 s relaxation delay with 1.59 s of acquisition time. Each FID (either relating to urine or to blood plasma) was zero-filled to 64k points and multiplied by a 0.3 Hz exponential line-broadening function prior to Fourier transformation. Spectra were manually phased and baseline corrected and chemical shifts referenced internally to TSP (at δ = 0.0 ppm) for urine and to R-glucose H1 resonance (at δ = 5.23 ppm) for plasma (where no TSP was added). All peak assignments were carried out with basis on 2D NMR experiments and consultation of the Bruker Biorefcode spectral database, as well as of other existing databases23,24 and specific compound standard solutions. Chemometrics

For the analysis of urine spectra, each spectrum was divided into “buckets” of variable width (from 0.005 to 0.03 ppm) to minimize first derivative-like effects derived from small peak shifts (e.g., pH dependent). The water resonance region (4.65 5.09 ppm) was excluded, as well as that of urea (5.50 6.20 ppm). For plasma spectral analysis, each set of spectra (standard, T2-edited and diffusion-edited) was used to construct data matrices for analysis, both using the full resolution spectra and spectra divided into “buckets” of fixed size (0.01 ppm). The water resonance region was excluded, as well as those of ethanol (1.15 1.20 and 3.62 3.68 ppm), which was found 3734

dx.doi.org/10.1021/pr200352m |J. Proteome Res. 2011, 10, 3732–3742

Journal of Proteome Research

ARTICLE

Figure 2. PLS-DA scores plots (left) and corresponding loadings (right) obtained for (a) standard 1H NMR spectra of urine (LV = 2) of the FM group (Δ, n = 26) vs controls (2, n = 25) and (b) T2-edited 1H NMR spectra of blood plasma (LV =3) of the FM group (Δ, n = 27) vs controls (2, n = 20). Loadings are colored according to Variable Importance to the Projection (VIP) and some assignments are indicated.

randomly in blood samples, possibly due to disinfection before blood collection. All spectra were normalized by dividing each spectral point/bucket by the total spectral area, to account for differences in biofluid volume. Principal component analysis (PCA),25 partial least-squares discriminant analysis (PLS-DA)26 and its orthogonal variant, O-PLS-DA,27 were applied to the different variable scaling methods for comparison purposes (unit variance, Pareto and  centered scaling), using SIMCA-P 11.5 (Umetrics, Umea, Sweden) software. The results shown below refer to those obtained using unit variance scaling, found to give the best results (e.g., as given by Q2 values distribution). The loadings corresponding to the PCA, PLS-DA, and OPLS-DA models obtained were back-transformed by multiplying all values by their standard deviation. Relevant peaks, identified in loadings profiles, were integrated using Amix 3.9.5, BrukerBioSpin, Rheinstetten, Germany, and analyzed using standard univariate analysis tests: Shapiro-Wilk normality test, Student’s t test or Wilcoxon test (respectively for normally or non-normally distributed data). All statistical tests and boxplots were performed using R-statistical software.28 The latter was also used, along with the Plotrix package,29 to produce PLS-DA and OPLS-DA loadings plots color-coded as a function of variable importance in the projection (VIP). Monte Carlo cross-validation, MCCV (using a partition of 7 blocks for each iteration), was applied to the main data sets, using 500 runs, with recovery of Q2 values, classification rates (by means of confusion matrices) and model complexity distribution. Moreover, a permutation strategy was used within the MC simulation, in order to assess each model’s prediction power,30,31 followed by examination of the permuted model parameter distribution and comparison to the original models (best models showing no or less overlap between permuted and original models). Statistical total correlation spectroscopy (STOCSY)32 was performed for all unassigned resonances, using R 2.9.2

software, to identify high positive correlations, possibly arising from signals from the same spin system.

’ RESULTS Figure 1 shows representative spectra of urine (left) and plasma (right) from a healthy pregnant woman, in her second trimester of pregnancy. Since no significant amounts of macromolecules are expected in urine, only standard NMR spectra were obtained in this work, as reflecting the overall metabolic profile of the biofluid. For blood plasma, however, the three types of spectra—standard, T2-edited and diffusion-edited—have been recorded, enabling a clearer separation of the signals arising from low-Mw metabolites, in the T2-edited spectra (Figure 1b, middle) from the macromolecules (mainly lipoproteins) that define the typical spectral profile of plasma diffusion-edited spectra (Figure 1b, bottom). A systematic study of spectral profile changes for different disorders, compared to controls, was carried out considering all types of spectra, several different scaling methods (unit variance, Pareto and centered) and a number of multivariate analysis methods (PCA, PLS-DA and OPLS-DA). The results shown below correspond to the best combination of experiment type, scaling, and multivariate methods and a more complete account of the results is shown in Table S1 (Supporting Information). Figure 2 shows the PLS-DA scores scatter plots obtained respectively for urine (Figure 2a) and blood plasma (Figure 2b) of pregnant women carrying malformed fetuses, compared to controls. It seems clear that the FM group is separated from the controls cloud in both biofluids (with Q2 0.31 and 0.48, respectively), suggesting a relatively high impact of these disorders on urine and plasma composition, consistently with previous reports on amniotic fluid composition.10 Also following earlier results,9 no consistent trends were observed for different 3735

dx.doi.org/10.1021/pr200352m |J. Proteome Res. 2011, 10, 3732–3742

Journal of Proteome Research

ARTICLE

Figure 3. PLS-DA scores scatter plot obtained for T2-edited 1H NMR spectra of blood plasma of women with fetuses with malformation of the central nervous system (CNS) (0, n = 9) (a) vs controls (9, n = 20) (LV = 3) and (b) vs other malformations ([, n = 19) (LV = 2).

gestational ages (e.g., taken as subgroups with >20 g.w. and 30% and should be taken only as qualitative tendencies. d Significance level of 95% was considered (p < 0.05). e Compound change not detected by PLS-DA (spectral region removed due to random changes of neighboring ethanol peak, possibly due to sample contamination) but confirmed by spectral integration. f Tentative assignment.

the third sample showing abnormally high lysine and carnitine resonances (no clinical explanations were found for these observations). Although the increase in glucose in the former two samples may relate to developing GDM, these samples must be considered as outliers at this stage, since they do not represent the whole sample group. Similarly, the plasma model in Figure 4b was obtained upon removal of one outlier (a control sample with higher glucose level). The low Q2 values (0.23 and 0.28, respectively for urine and plasma) and large superposition of Q2 values distributions (original and permuted models, data not shown), mainly for plasma data, lead to classification rates not

higher than 75%, thus reflecting a relatively low predictive ability of the models. Regarding the urine model (Figure 4a), it is clear that the bulk of the groups separate along LV1, with one of the prediagnostic GDM samples clearly mixed in the controls (overweight subject, BMI 26.7, suffering from hypertension) and an apparent subgroup of 4 samples clustering toward positive LV1. In spite of the low predictive power of this model, analysis of the corresponding loading weights (w[1]) (not shown) and subsequent spectral inspection and integration indicated significant increases in: 3-hydroxyisovalerate, 2-hydroxyisobutyrate, choline (with p-value 4.87  10 5), 2PY, NMND and unassigned 2 3737

dx.doi.org/10.1021/pr200352m |J. Proteome Res. 2011, 10, 3732–3742

Journal of Proteome Research

ARTICLE

Figure 4. (a) PLS-DA scores scatter plot for standard 1H NMR spectra of urine (LV = 2) of prediagnostic GDM subjects (0, n = 26) vs controls (9, n = 25); (b) PLS-DA scores scatter plot for T2-edited 1H NMR spectra of blood plasma (LV = 2) of prediagnostic GDM subjects (0, n = 14) vs controls (9, n = 19); (c) OPLS-DA scores for standard 1H NMR spectra of urine (LV = 1 + 2) of women carrying fetuses with chromosomal disorders (O, n = 22) vs controls (b, n = 25); (d) OPLS-DA scores for the T2-edited 1H NMR spectra of blood plasma (LV = 1 + 1) of women carrying fetuses with chromosomal disorders (O, n = 23) vs controls (b, n = 20); (e) OPLS-DA scores for standard 1H NMR spectra of urine (LV = 1 + 1) of pre-PTD subjects (), n = 15) vs controls ((, n = 25).

(1.1/4.10 ppm) and 4 (7.68 ppm) (Table 2). Boxplots for choline, 2PY and NMND are shown in Figure S2b (Supporting Information). Regarding blood plasma, the group separation observed in Figure 4b was found to relate to decreasing tendencies in TMAO and betaine, as also expressed by the corresponding boxplots (Figure S3b, Supporting Information). In relation to chromosomal disorders, some impact is apparent on urine (Q2 = 0.36) (Figure 4c), along with a subtle effect on blood plasma (Q2 = 0.24) (Figure 4d), although Q2 values

distribution (permuted and original) were found considerably superimposed for both data sets (data not shown) and weak average classification powers (62 67%) were found. The urine model was obtained after visually identifying an outlier sample (found to correspond to a subject affected by a thyroid disorder), whereas the plasma model involved the removal of a sample with significantly low lipids content (but otherwise normal clinical record). As indicated in Table 2, urine shows increasing tendencies in choline and unassigned 2 (1.1 ppm). Furthermore, urine spectra 3738

dx.doi.org/10.1021/pr200352m |J. Proteome Res. 2011, 10, 3732–3742

Journal of Proteome Research

ARTICLE

(96%) of permuted models showed lower Q2 values than the original ones (data not shown). The loadings weights (w[1]) analysis indicated higher levels of choline and unassigned 2 (1.1/ 4.10 ppm), together with 2-hydroxyisobutyrate in pre-PTD samples. In the model shown, two outlier samples were removed corresponding to high glucose levels. Finally, for the pre-PROM group, no changes were detected in urine and, for plasma, only faint alterations were suggested in some samples for acetate, glutamine, citrate and albumin (results not shown).

’ DISCUSSION Fetal Malformations Group

Figure 5. (a) PLS-DA scores scatter plot for T2-edited 1H NMR spectra of blood plasma of women carrying fetuses with trisomy 21 (Δ, n = 8) vs controls (9, n = 20) (LV = 2) and (b) Expansion of representative T2edited 1H NMR spectra of controls (left) and trisomy 21 samples (right), showing profile differences in the lipids CH3/(CH2)n region.

clearly reflect a different profile for trisomy 21 samples, compared to the heterogeneous group of the remaining chromosomal disorders (Figure S4, Supporting Information), expressing changed levels (with p < 0.05) for several unassigned signals (1.42 ppm, d; 1.82 ppm, d; 4.40 ppm, s) and identifying urine as potentially sensitive to chromosomal disorder type. Regarding plasma, the loading weights (w[1]) (not shown) corresponding to the scores obtained for all samples (Figure 4d) suggested small qualitative changes in lipid profiles, difficult to confirm through signal integration. Interestingly, changes in lipid signals were seen to gain additional relevance when considering trisomy 21 samples alone vs controls, for which an improved PLS-DA scores scatter plot was obtained (Figure 5a). In spite of the low number of trisomy 21 plasma samples (n = 8), it may be seen (Figure 5b) that, apart from the sample indicated as T8, the methyl band at 0.85 ppm shows a tendency for broader and more structured profiles. Also, the CH2 band at ca. 1.26 ppm shows a narrowing tendency for some of the trisomy 21 samples. The above results show that group heterogeneity in terms of chromosomal disorder type seem to affect both urine (Figure S4, Supporting Information) and plasma (Figure 5) compositions, and that lipids seem to be a compound family of interest for further targeted analysis. For the subjects who delivered prematurely, the pre-PTD group, only urine could be analyzed because an insufficient number of plasma samples were collected (n = 4). Figure 4e shows the PLS-DA scores scatter plot (Q2 = 0.33) obtained, this model having shown to have predictive power as the majority

Maternal urine corresponding to the FM group reflects disturbances in amino acid metabolism, particularly branched chain amino acid metabolism, as reflected by the enhanced excretion of the glucogenic (and essential) amino acids valine and isoleucine. However, these were seen to decrease in the amniotic fluid of malformed fetuses, possibly due to enhanced gluconeogenesis under hypoxia conditions.9,10 This expresses an apparent contradiction between fetal and mother’s needs, since the enhanced fetal need for these amino acids (which should result in the shifting of the mother’s amino acid supplies through the placenta) somehow leads to their higher excretion in maternal urine, thus raising the question of how directly fetal metabolic disorders reflect onto the mother’s metabolism, an issue that to our knowledge remains somewhat unclear. Threonine is seen to increase in maternal urine, as well as in amniotic fluid.10 The underuse of this glucogenic/ketogenic amino acid in gluconeogenesis has been suggested for malformed fetuses, together with its possible involvement in other pathways.10 Cisaconitate, an intermediate in the tricarboxylic acid (TCA) cycle, showed enhanced levels in FM urine suggesting a possible higher demand on the TCA cycle, probably as a result of the increased fetal need for glucose production through gluconeogenesis. Another variation, apparently specific to FM maternal urine, is the increase in hypoxanthine. This may be a sign of rapid ATP degradation under fetal hypoxia conditions, as noted previously in urine in an animal model study34 and in plasma, in relation to oxidative kidney damage due to perinatal hypoxia in preterm newborns.35 However, this compound was not found elevated in amniotic fluid.10 Other urinary effects observed here, such as increases in choline, 2PY and NMND, seem to be less specific of disorder type, the former occurring for all disorders under study (apparently in tandem with increased Un 2 at 1.1/4.10 ppm). The latter two compounds also show increases in the prediagnostic GDM group (Table 2). The above observations suggest that this set of compounds probably arise as a result of the general effects of stress (the nature of which is unclear at this stage), rather than from a specific disease outcome. Choline has a known relationship to fetal brain development and its metabolism intersects with methionine and folate metabolisms, at the point when homocysteine is converted into methionine, using up betaine.36,37 It is possible that a slowing down in homocysteine-methionine conversion occurs in FM cases, as noted before for fetal neural tube defects38 and cardiac defects,39 thus leading to higher choline levels excreted. Choline has also been shown to change in relation to hypoxia and pre-eclampsia,40 thus advancing the alternative hypothesis of a possible underlying hypoxia condition. Finally, the changes in 2PY and NMND urinary levels reflect disturbances in the nucleotide metabolism, and in 3739

dx.doi.org/10.1021/pr200352m |J. Proteome Res. 2011, 10, 3732–3742

Journal of Proteome Research particular for NMND, in the tryptophan-NAD+ pathway, as suggested for type 2 diabetic subjects41 and autistic subjects.42 Some of the changes observed in maternal plasma relating to the FM group are consistent with the proposals advanced above. First, the increases in acetone and 3-hydroxybutyric acid should result from storage lipids conversion into ketone bodies and transfer through blood for TCA enhancement and energy production in extra-hepatic tissues. This is consistent with the proposed fetal state of glucose depletion and subsequent gluconeogenesis enhancement.43 Regarding the remaining ketone body, acetoacetate, this is indeed detected in the NMR spectrum but was not seen to change significantly between controls and FM groups. Furthermore, the small increase in plasma lipids, of undetermined nature at this stage, may be a reflection of the triglyceride conversion process mentioned above. The lower level of betaine in plasma may possibly relate to the above-noted disturbance in the homocysteine-methionine conversion, which has also reflections in urinary choline levels. At this stage, no explanation is available in relation to the methanol decrease in plasma, although its endogenous nature has been noted and discussed before.44 Regarding the small spectral changes seen for CNS malformations (Figure 3), it may be that lipid metabolism is affected in different manners, depending on the type of malformation, thus justifying further analysis on this metabolite family. Prediagnostic GDM Group

Prediagnostic GDM subjects show less but yet measurable effects on maternal urine. The increases in 3-hydroxyisovalerate and 2-hydroxyisobutyrate seem to be specific of this group, while choline (and Un 2), 2PY and NMND may arise from common stress effects, as suggested above for the FM group. Regarding the increased excretion of 3-hydroxyisovalerate, it has been suggested that in pregnancy this compound may reflect a reduced biotin status, probably related to the decreased activity of the biotin-dependent enzyme methylcrotonyl-CoA.45 In this context, biotin deficiency has been suggested to relate to certain human birth defects but, to our knowledge, no report of a relation to the prediagnostic GDM condition exists. In relation to 2-hydroxyisobutyrate, increased urinary excretion of this compound has been reported in type 2 diabetic mice, and this variation was correlated with the degradation pathway of valine, leucine and isoleucine.46 In addition, urinary excretion of this compound has also been seen in obese subjects and suggested to derive from the microbial degradation of proteins that escape digestion in the upper gastrointestinal tract.47 Out of the 26 women considered in the prediagnostic GDM group (after removal of 3 outliers), 13 had BMI > 25 (overweight), whereas 7 had BMI < 25 and BMI information was unavailable for 6; therefore, a relationship between high 2-hydroxyisobutyrate excretion and high BMI may not be ruled out. Regarding plasma composition, decreases in TMAO and betaine were noted. The latter, again, may reflect some changes affecting choline-related homocysteine-methionine conversion36,37 but, to our knowledge, no relation has been reported for the GDM condition. However, in relation to diabetes, higher urinary betaine levels have been found in diabetic subjects48 whereas lower betaine levels have been found in the plasma of subjects suffering from metabolic syndrome, which comprises a number of different disorders, including diabetes.49 Regarding TMAO, the reason for its increase is unclear at this stage, one possibility being a relation to kidney function,50 although a diet-related origin (specifically involving the ingestion of fish) may not be ruled out.

ARTICLE

Chromosomal Disorders and Pre-PTD Groups

Chromosomal disorder cases seem to be generally characterized by an increase in choline (and Un2) in urine, as a result of imposed fetal stress, as suggested above. Since no tandem change in plasma betaine is noted (contrary to the observations for the previous groups), it is possible that this choline change has its origin in distinct metabolic, for example, hypoxia effects40 and/or in relation to lipid metabolism. Indeed, significant qualitative changes in plasma lipids (for which no further assignment may be advanced at this stage) seem to be at the root of the weak separation of the chromosomal disorders group from controls. In relation to disorder type, urine seems to show relatively high sensitivity, although based on unassigned NMR features. In plasma, some sensitivity is also noted to chromosomal disorder type, mainly based on lipid changes. In this context, it is important to note the tendency for higher BMI values characterizing the subjects carrying fetuses with chromosomal disorders (Table 1), although this information is only known for a subgroup of subjects. The relevance of high BMI values in this subgroup of subjects should thus be thoroughly investigated in future studies. Pre-PTD maternal urine again shows signs of choline metabolism perturbation, in tandem with an increasing tendency in 2-hydroxyisobutyrate, the latter possibly relating to amino acid metabolism. In this case, BMI information is only available for a few samples, so that any relation to high BMI values requires further investigation. To our knowledge, no reports of these compounds changing have been made in relation to PTD cases.

’ CONCLUSIONS This work shows that second trimester maternal urine and plasma may be a useful window for detecting and monitoring metabolic changes that accompany prenatal disorders. Among the disorders hereby studied, fetal malformations are shown to have a strong impact on the metabolite composition of both urine and plasma. Some metabolites seem to play a specific role in distinguishing FM from controls—glucogenic amino acids, cisaconitate, acetone, 3-hydroxybutyric, hypoxanthine—their changes supporting earlier suggestions that malformed fetuses seem to be affected by hypoxia (in agreement with hypoxanthine changes), leading to enhanced anaerobic glucose usage. This subsequently justifies a need for glucose replenishment through production of ketone bodies (as noted through acetone and 3-hydroxybutyric acid increases in plasma), along with higher demand on the TCA cycle (suggested by overproduction of glucogenic amino acis and cis-aconitate), to enable gluconeogenesis. Choline, betaine, 2PY and NMND levels also change significantly but apparently also reflecting more general stress effects, the precise nature of which is, to our knowledge, unclear at this stage. In FM, altered choline metabolism may result from hypoxia and/or perturbations in homocysteine-methionine conversion, as noted by tandem choline (in urine) and betaine (in plasma) variations, and 2PY and NMND changes suggest nucleotide metabolism disturbances. In the case of different FM types (e.g., CNS malformations, compared to others), plasma seems to exhibit higher sensitivity to malformation type, based mainly on lipid changes. All remaining disorders reflect smaller metabolite changes in both urine and plasma, with subjects in prediagnostic GDM conditions exhibiting apparently specific changes in 3-hydroxyisovalerate and 2-hydroxyisobutyrate, which suggest, respectively, early changes in biotin status and altered amino acid and/or gut metabolisms (the latter possibly relating to the higher BMI values 3740

dx.doi.org/10.1021/pr200352m |J. Proteome Res. 2011, 10, 3732–3742

Journal of Proteome Research of a significant number of prediagnostic GDM subjects). Again, choline and nucleotide metabolic disturbances seem to take place. Finally, chromosomal disorders and pre-PTD groups were shown to undergo changes in choline and 2-hydroxyisobutyrate, for which additional work is necessary to attempt further interpretation. In the case of chromosomal disorders, both urine and plasma seem to enable distinction of trisomy 21 cases from other cases. No relevant changes were observed for pre-PROM subjects. The above results show the promise of second trimester maternal biofluid metabonomics to identify the metabolic changes underlying prenatal disorders, although statistical improvement of the models is required (for instance, through increased sample numbers) for identification of sturdy diseasespecific biomarkers to be envisaged. Also, the biochemical interpretations proposed here require further and thorough demonstration in order for suitable understanding of metabolic fetus/mother synergism to be achieved. However, the results shown here already unveil the large potential of urine and plasma metabolite markers for clinical use as prenatal diagnostic and prognostic tools. Further advancements may entail a more global approach, possibly using both NMR and MS, for multiple biofluid analysis as well as placenta analysis, the latter having been initiated in recent years.40,51 The tandem use of this information with existing clinical methods is likely to lead to significant improvements in pregnancy monitoring and outcome.

’ ASSOCIATED CONTENT

bS

Supporting Information Table S1 and Figures S1 S4. This material is available free of charge via the Internet at http://pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*Tel +351 234 370707, fax +351 234 370084, e-mail [email protected]. Author Contributions †

These authors have contributed equally to this work.

’ ACKNOWLEDGMENT We acknowledge the Foundation for Science and Technology (FCT), Portugal, for funding through research project PTDC/ QUI/66523/2006 and grants SFRH/BD/41869/2007, SFRH/ BD/64159/2009 and SFRH/BD/73343/2010 for G.G., S.D., and J.P., respectively. The Portuguese National NMR Network (RNRMN), supported with FCT funds is also acknowledged and we are grateful to M. Spraul, Bruker BioSpin, Germany, for providing access to spectral databases. ’ REFERENCES (1) Hourrier, S.; Salomon, L. J.; Dreux, S.; Muller, F. Screening for adverse pregnancy outcome at early gestational age. Clin. Chim. Acta 2010, 411 (21 22), 1547–1552. (2) Hacker, N. F.; Moore, J. G.; Gambone, J. C. Essentials of Obstetrics and Gynecology, 4th ed.; Elsevier: New York, 2004. (3) Nicholson, J. K.; Lindon, J. C.; Holmes, E. 'Metabonomics’: understanding the metabolic responses of living systems to pathophysiological stimuli via multivariate statistical analysis of biological NMR spectroscopic data. Xenobiotica 1999, 29 (11), 1181–1189.

ARTICLE

(4) Bock, J. Metabolic profiling of amniotic fluid by proton nuclear magnetic resonance spectroscopy: correlation with fetal maturation and other clinical variables. Clin. Chem. 1994, 40 (1), 56–61. (5) Joe, B. N.; Vahidi, K.; Zektzer, A.; Chen, M.-H.; Clifton, M. S.; Butler, T.; Keshari, K.; Kurhanewicz, J.; Coakley, F.; Swanson, M. G. 1H HR-MAS spectroscopy for quantitative measurement of choline concentration in amniotic fluid as a marker of fetal lung maturity: Inter- and intraobserver reproducibility study. J. Magn. Reson. Imaging 2008, 28 (6), 1540–1545. (6) Sims, C. J.; Fujito, D. T.; Burholt, D. R.; Dadok, J.; Wilkinson, D. A. Comparison of metabolite levels in second and third trimester human amniotic fluid samples using proton magnetic resonance spectroscopy. J. Matern.-Fetal Invest. 1996, 6 (2), 62–66. (7) Cohn, B. R.; Fukuchi, E. Y.; Joe, B. N.; Swanson, M. G.; Kurhanewicz, J.; Yu, J. W.; Caughey, A. B. Calculation of gestational age in late second and third trimesters by ex vivo magnetic resonance spectroscopy of amniotic fluid. Am. J. Obstet. Gynecol. 2010, 203 (1), 76. e1–76.e10. (8) Groenen, P. M. W.; Engelke, U. F.; Wevers, R. A.; Hendriks, J. C. M.; Eskes, T. K. A. B.; Merkus, H. M. W. M.; Steegers-Theunissen, R. P. M. High-resolution 1H NMR spectroscopy of amniotic fluids from spina bifida fetuses and controls. Eur. J. Obstet. Gynecol. Reprod. Biol. 2004, 112 (1), 16–23. (9) Grac) a, G.; Duarte, I. F.; Barros, A. S.; Goodfellow, B. J.; Diaz, S.; Carreira, I. M.; Couceiro, A. B.; Galhano, E.; Gil, A. M. 1H NMR based metabonomics of human amniotic fluid for the metabolic characterization of fetus malformations. J. Proteome Res. 2009, 8 (8), 4144–4150. (10) Grac) a, G.; Duarte, I. F.; Barros, A. S.; Goodfellow, B. J.; Diaz, S. O.; Pinto, J.; Carreira, I. M.; Galhano, E.; Pita, C.; Gil, A. M. Impact of prenatal disorders on the metabolic profile of second trimester amniotic fluid: a nuclear magnetic resonance metabonomic study. J. Proteome Res. 2010, 9 (11), 6016–6024. (11) Romero, R.; Mazaki-Tovi, S.; Vaisbuch, E.; Kusanovic, J. P.; Chaiworapongsa, T.; Gomez, R.; Nien, J. K.; Yoon, B. H.; Mazor, M.; Luo, J. Q.; Banks, D.; Ryals, J.; Beecher, C. Metabolomics in premature labor: a novel approach to identify patients at risk for preterm delivery. J. Matern.-Fetal Neonat. Med. 2010, 23 (12), 1344–1359. (12) Ottolenghi, C.; Abermil, N.; Lescoat, A.; Aupetit, J.; Beaugendre, O.; Morichon-Delvallez, N.; Ricquier, D.; Chadefaux-Vekemans, B.; Rabier, D. Gestational age-related reference values for amniotic fluid organic acids. Prenatal Diagn. 2010, 30 (1), 43–48. (13) Amaral, C.; Gallardo, E.; Rodrigues, R.; Pinto Leite, R.; Quelhas, D.; Tomaz, C.; Cardoso, M. L. Quantitative analysis of five sterols in amniotic fluid by GC-MS: Application to the diagnosis of cholesterol biosynthesis defects. J. Chromatogr., B 2010, 878 (23), 2130–2136. (14) Georgiou, H. M.; Lappas, M.; Georgiou, G. M.; Marita, A.; Bryant, V. J.; Hiscock, R.; Permezel, M.; Khalil, Z.; Rice, G. E. Screening for biomarkers predictive of gestational diabetes mellitus. Acta Diabetol. 2008, 45 (3), 157–165. (15) Sanchez-Vera, I.; Bonet, B.; Viana, M.; Quintanar, A.; Martin, M. D.; Blanco, P.; Donnay, S.; Albi, M. Changes in plasma lipids and increased low-density lipoprotein susceptibility to oxidation in pregnancies complicated by gestational diabetes: consequences of obesity. Metab.-Clin. Exp. 2007, 56 (11), 1527–1533. (16) Jelliffe-Pawlowski, L. L.; Baer, R. J.; Currier, R. J. Second trimester serum predictors of preterm birth in a population-based sample of low-risk pregnancies. Prenatal Diagn. 2010, 30 (8), 727–733. (17) Turner, E.; Brewster, J. A.; Simpson, N. A. B.; Walker, J. J.; Fisher, J. Plasma from women with Preeclampsia has a low lipid and ketone body content - A nuclear magnetic resonance study. Hypertens. Pregnancy 2007, 26 (3), 329–342. (18) Turner, E.; Brewster, J. A.; Simpson, N. A. B.; Walker, J. J.; Fisher, J. Aromatic amino acid biomarkers of preeclampsia - A nuclear magnetic resonance investigation. Hypertens. Pregnancy 2008, 27 (3), 225–235. (19) Kenny, L. C.; Broadhurst, D. I.; Dunn, W.; Brown, M.; North, R. A.; McCowan, L.; Roberts, C.; Cooper, G. J. S.; Kell, D. B.; Baker, 3741

dx.doi.org/10.1021/pr200352m |J. Proteome Res. 2011, 10, 3732–3742

Journal of Proteome Research P. N. Robust early pregnancy prediction of later preeclampsia using metabolomic biomarkers. Hypertension 2010, 56 (4), 741–749. (20) Scioscia, M.; Gumaa, K.; Selvaggi, L. E.; Rodeck, C. H.; Rademacher, T. W. Increased inositol phosphoglycan P-type in the second trimester in pregnant women with type 2 and gestational diabetes mellitus. J. Perinat. Med. 2009, 37 (5), 469–471. (21) Marcos, J.; Craig, W. Y.; Palomaki, G. E.; Kloza, E. M.; Haddow, J. E.; Roberson, M.; Bradley, L. A.; Shackleton, C. H. L. Maternal urine and serum steroid measurements to identify steroid sulfatase deficiency (STSD) in second trimester pregnancies. Prenatal Diagn. 2009, 29 (8), 771–780. (22) Stein, T. P.; Scholl, T. O.; Schluter, M. D.; Leskiw, M. J.; Chen, X. H.; Spur, B. W.; Rodriguez, A. Oxidative stress early in pregnancy and pregnancy outcome. Free Radical Res. 2008, 42 (10), 841–848. (23) Wishart, D. S.; Tzur, D.; Knox, C.; Eisner, R.; Guo, A. C.; Young, N.; Cheng, D.; Jewell, K.; Arndt, D.; Sawhney, S.; Fung, C.; Nikolai, L.; Lewis, M.; Coutouly, M.-A.; Forsythe, I.; Tang, P.; Shrivastava, S.; Jeroncic, K.; Stothard, P.; Amegbey, G.; Block, D.; Hau, D. D.; Wagner, J.; Miniaci, J.; Clements, M.; Gebremedhin, M.; Guo, N.; Zhang, Y.; Duggan, G. E.; MacInnis, G. D.; Weljie, A. M.; Dowlatabadi, R.; Bamforth, F.; Clive, D.; Greiner, R.; Li, L.; Marrie, T.; Sykes, B. D.; Vogel, H. J.; Querengesser, L. HMDB: the Human Metabolome Database. Nucleic Acids Res. 2007, 35 (suppl 1), D521–D526. (24) Seavey, B. R.; Farr, E. A.; Westler, W. M.; Markley, J. L. J. Biomol. NMR 1991, 1, 217–236http://www.bmrb.wisc.edu/metabolomics/metabolomics_standards.html. (25) Jolliffe, I. T. Principal Component Analysis, 2nd ed.; Springer: New York, 2002. (26) Barker, M.; Rayens, W. Partial Least Squares for discrimination. J. Chemom. 2003, 17 (3), 166–173. (27) Trygg, J.; Wold, S. Orthogonal projections to latent structures (O-PLS). J. Chemom. 2002, 16 (3), 119–128. (28) R Development Core Team. R: A Language and Environment for Statistical Computing, 2.9.2; R Foundation for Statistical Computing: Vienna, Austria, 2010. (29) Lemon, J. Plotrix: a package in the red light district of R. R-News 2006, 6 (4), 8–12. (30) Wiklund, S.; Nilsson, D.; Eriksson, L.; Sj€ostr€om, M.; Wold, S.; Faber, K. A randomization test for PLS component selection. J. Chemom. 2007, 21 (10 11), 427–439. (31) Westerhuis, J. A.; Hoefsloot, H. C. J.; Smit, S.; Vis, D. J.; Smilde, A. K.; Van Velzen, E. J. J.; Van Duijnhoven, F. A. D. Assessment of PLSDA cross validation. Metabolomics 2008, 4 (1), 81–89. (32) Cloarec, O.; Dumas, M.-E.; Craig, A.; Barton, R. H.; Trygg, J.; Hudson, J.; Blancher, C.; Gauguier, D.; Lindon, J. C.; Holmes, E.; Nicholson, J. Statistical total correlation spectroscopy: an exploratory approach for latent biomarker identification from metabolic 1H NMR data sets. Anal. Chem. 2005, 77 (5), 1282–1289. (33) Brereton, R. G. Consequences of sample size, variable selection, and model validation and optimization, for predicting classification ability from analytical data. Trends Anal. Chem. 2006, 25 (11), 1103–1111. (34) Walker, V.; Bennet, L.; Mills, G. A.; Green, L. R.; Gnanakumaran, K.; Hanson, M. A. Effects of hypoxia on urinary organic acid and hypoxanthine excretion in fetal sheep. Pediatr. Res. 1996, 40 (2), 309–318. (35) Perrone, S.; Mussap, M.; Longini, M.; Fanos, V.; Bellieni, C. V.; Proietti, F.; Cataldi, L.; Buonocore, G. Oxidative kidney damage in preterm newborns during perinatal period. Clin. Biochem. 2007, 40 (9 10), 656–660. (36) Zeisel, S. H. Choline: critical role during fetal development and dietary requirements in adults. Annu. Rev. Nutr. 2006, 26 (1), 229–250. (37) Miller, A. L.; Kelly, G. S. Methionine and homocysteine metabolism and the nutritional prevention of certain birth defects and complications of pregnancy. Altern. Med. Rev. 1996, 4 (1), 220–235. (38) Zhao, W.; Mosley, B. S.; Cleves, M. A.; Melnyk, S.; James, S. J.; Hobbs, C. A. Neural tube defects and maternal biomarkers of folate, homocysteine, and glutathione metabolism. Birth Defects Res., Part A 2006, 76 (4), 230–236.

ARTICLE

(39) Hobbs, C. A.; Cleves, M. A.; Zhao, W.; Melnyk, S.; James, S. J. Congenital heart defects and maternal biomarkers of oxidative stress. Am. J. Clin. Nutr. 2005, 82 (3), 598–604. (40) Dunn, W. B.; Brown, M.; Worton, S. A.; Crocker, I. P.; Broadhurst, D.; Horgan, R.; Kenny, L. C.; Baker, P. N.; Kell, D. B.; Heazell, A. E. P. Changes in the metabolic footprint of placental explant-conditioned culture medium identifies metabolic disturbances related to hypoxia and preeclampsia. Placenta 2009, 30 (11), 974–980. (41) Salek, R. M.; Maguire, M. L.; Bentley, E.; Rubtsov, D. V.; Hough, T.; Cheeseman, M.; Nunez, D.; Sweatman, B. C.; Haselden, J. N.; Cox, R. D.; Connor, S. C.; Griffin, J. L. A metabolomic comparison of urinary changes in type 2 diabetes in mouse, rat, and human. Physiol. Genomics 2007, 29 (2), 99–108. (42) Yap, I. K. S.; Angley, M.; Veselkov, K. A.; Holmes, E.; Lindon, J. C.; Nicholson, J. K. Urinary metabolic phenotyping differentiates children with autism from their unaffected siblings and age-matched controls. J. Proteome Res. 2010, 9 (6), 2996–3004. (43) Herrera, E. Implications of dietary fatty acids during pregnancy on placental, fetal and postnatal development - a review. Placenta 2002, 23, S9–S19. (44) Haffner, H. T.; Graw, M.; Besserer, K.; Blickle, U.; Henge, C. Endogenous methanol: variability in concentration and rate of production. Evidence of a deep compartment?. Forensic Sci. Int. 1996, 79 (2), 145–154. (45) Mock, D. M. Marginal Biotin Deficiency is Common in Normal Human Pregnancy and Is Highly Teratogenic in Mice. J. Nutr. 2009, 139 (1), 154–157. (46) Connor, S. C.; Hansen, M. K.; Corner, A.; Smith, R. F.; Ryan, T. E. Integration of metabolomics and transcriptomics data to aid biomarker discovery in type 2 diabetes. Mol. BioSystems 2010, 6 (5), 909–921. (47) Calvani, R.; Miccheli, A.; Capuani, G.; Miccheli, A. T.; Puccetti, C.; Delfini, M.; Iaconelli, A.; Nanni, G.; Mingrone, G. Gut microbiomederived metabolites characterize a peculiar obese urinary metabotype. Int. J. Obes. 2010, 34 (6), 1095–1098. (48) Dellow, W. J.; Chambers, S. T.; Lever, M.; Lunt, H.; Robson, R. A. Elevated glycine betaine excretion in diabetes mellitus patients is associated with proximal tubular dysfunction and hyperglycemia. Diabetes Res. Clin. Pract. 1999, 43 (2), 91–99.  (49) Konstantinova, S. V.; Tell, G. S.; Vollset, S. E.; Nygard, O.; ^ Bleie, U.; Ueland, P. M. Divergent associations of plasma choline and betaine with components of metabolic syndrome in middle age and elderly men and women. J. Nutr. 2008, 138 (5), 914–920. (50) Fujiwara, M.; Kobayashi, T.; Jomori, T.; Maruyama, Y.; Oka, Y.; Sekino, H.; Imai, Y.; Takeuchi, K. Pattern recognition analysis for 1H NMR spectra of plasma from hemodialysis patients. Anal. Bioanal. Chem. 2009, 394 (6), 1655–1660. (51) Horgan, R. P.; Broadhurst, D. I.; Dunn, W. B.; Brown, M.; Heazell, A. E. P.; Kell, D. B.; Baker, P. N.; Kenny, L. C. Changes in the metabolic footprint of placental explant-conditioned medium cultured in different oxygen tensions from placentas of small for gestational age and normal pregnancies. Placenta 2010, 31 (10), 893–901.

3742

dx.doi.org/10.1021/pr200352m |J. Proteome Res. 2011, 10, 3732–3742