Newborn Meconium and Urinary Metabolome ... - ACS Publications

Subscriber access provided by UNIV LAVAL

Article

Newborn Meconium and Urinary Metabolome Response to Maternal Gestational Diabetes Mellitus: A Preliminary Case-Control Study Siyuan Peng, Jie Zhang, Liangpo Liu, Xueqin Zhang, Qingyu Huang, Ambreen Alamdar, Meiping Tian, and Heqing Shen J. Proteome Res., Just Accepted Manuscript • DOI: 10.1021/pr5011857 • Publication Date (Web): 27 Feb 2015 Downloaded from http://pubs.acs.org on March 4, 2015

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Journal of Proteome Research is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 41

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Journal of Proteome Research

Newborn Meconium and Urinary Metabolome Response to Maternal Gestational Diabetes Mellitus: A Preliminary Case-Control Study Siyuan Peng†1, Jie Zhang†1, Liangpo Liu†, Xueqin Zhang‡, Qingyu Huang†, Ambreen Alamdar†, Meiping Tian†, Heqing Shen†,* †

Key Lab of Urban Environment and Health, Institute of Urban Environment, Chinese Academy of

Sciences, Xiamen, PR China, 361021; ‡

1

Xiamen Maternity and Child Health Care Hospital, Xiamen, PR China, 361003

The authors contribute equally;

*To whom correspondence may be addressed: Prof. Heqing Shen, Institute of Urban Environment, Chinese Academy of Sciences, 1799 Jimei Road, Xiamen, 361021, China; Tel/Fax: (86)-592-6190771; E-mail: [email protected]

Running title: Maternal GDM Altered Newborn Meconium and Urinary Metabolome

KEYWORDS Meconium, Urine, Newborn, Gestational diabetes mellitus, Metabolomics, Biomarkers

ACS Paragon Plus Environment 1 / 36


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ABSTRACT

Recently, the number of women suffering from gestational diabetes mellitus (GDM) dramatically increases. GDM attracts more and more attentions due to its potential harm to the heath of both the fetus and mother. We designed this case-control study to investigate the metabolome response of newborn meconium and urine to maternal GDM. GDM mothers (n=142) and healthy controls (n=197) were recruited during June to July, 2012 in Xiamen, China. The newborns’ metabolic profiles were acquired using liquid chromatography coupled to mass spectrometer. The data showed that meconium and urinary metabolome pattern clearly discriminated GDM cases from controls. Fourteen meconium metabolic biomarkers and three urinary metabolic biomarkers were tentatively identified for GDM. Altered levels of various endogenous biomarkers revealed that GDM may induce disruptions of lipid metabolism, amino acid metabolism and purine metabolism. An unbalanced lipid pattern is suspected to be a GDM-specific feature. Furthermore, the associations between the potential biomarkers and GDM risk were evaluated by binary logistic regression and receiver operating characteristic analysis. A combined model of nine-meconium biomarkers suggested a great potential in diagnosing GDM-induced disorders.


Page 2 of 41

Page 3 of 41

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


INTRODUCTION

Gestational diabetes mellitus (GDM) is a specific type of diabetes mellitus. It is defined as any degree of glucose intolerance with onset or first recognition during pregnancy.1 Women with GDM are more likely to develop type 2 diabetes mellitus (T2DM) in the years following pregnancy.2 Estimated GDM prevalence ranges from 1 to 14% in different countries.3 In China, GDM incidence increased from 2.4% in 1999 to 8.2% in 2012 for urban women ( ≥ 20 years old).4

Investigation of disrupted global metabolic status (metabolite profile) helps us improve the understanding of molecular mechanism of GDM processes. Metabolomics is capable of capturing disease-relevant metabolic profile changes and providing novel biomarkers of disease processes.5,6 It has also been used to investigate metabolic alterations of GDM women during pregnancy7. GDM women exhibited increased levels of 3-hydroxyisovalerate and 2-hydroxyisobutyrate in urine8 and decreased levels of lipid metabolism-related metabolites (esp. lysoglycerophospholipids) in plasma9. However, other studies failed to yield reliable novel biomarkers in GDM women’s urine and fetal amniotic fluid (AF).10,11 In general, published literatures regarding the GDM specific metabolic alterations are scare and they only focus on unique alterations of maternal biological fluid samples (plasma, urine) or fetal AF. These obtained inclusive findings remain to be improved.

GDM is proposed to induce metabolic dysfunction for both mothers and fetus due to fetal-maternal interaction during gestation.7 So far numbers of complications associated with mothers and their fetus have been well documented.12 Adverse maternal consequence of GDM is mainly postpartum risk of diabetes mellitus2 while more intrapartum risks are associated with fetus, including neonatal



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

macrosomia, hypoglycemia, perinatal mortality, congenital malformation and respiratory distress syndrome and so on.12 However, few GDM metabolomics investigation has been reported for the fetus. Matrix selection is a problem for fetal metabolome study, because the widely used matrix (i.e. maternal urine and blood) only reflect transient or recent condition on the maternal side. Instead, newborn meconium is a promising choice. Meconium is formed by the fetus as early as the 12th week of gestation and accumulates until birth, thus it is a repository of endogenous and exogenous metabolites that trans-placentally transfer from the mother to the fetus.13 Meconium provides a longer historical record for prenatal metabolites accumulation. Therefore, meconium metabolome is considered as a novel tool to detect metabolic alterations for both mothers and fetus during pregnancy. Besides meconium, fetal urine and AF may be useful matrix for monitoring gestational metabolic changes, but both of them are difficult to collect and their composition changes with gestational age.14 Given AF mainly consists of fetal urine at the end of gestation14, and newborn urine, which is excreted in the first postnatal day, may be very close to AF at delivery. Therefore, newborn first urine may also reflect the recent metabolic condition before delivery.

In this study, the newborn meconium and urine samples were collected from 142 GDM and 197 healthy subjects. A liquid chromatography/mass spectrometry (LC/MS) based metabolomics strategy was employed to acquire and discriminate the global newborn meconium and urinary metabolic profiles. To our knowledge, this is the first metabolomics study of newborn meconium and urine to probe the effects of maternal GDM. Several endogenous potential biomarkers were tentatively identified, which could indicate a long-term metabolic dysfunction for GDM mothers and their babies during gestation.


Page 4 of 41

Page 5 of 41

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


EXPERIMENTAL SECTION

Participant Recruitment and Sample Collection

Details of participants’ recruitment, GDM screening and case-control pairing have been described in elsewhere.15 Briefly, a total of 1359 pregnant women were recruited during June to July, 2012 in Xiamen Maternity and Child Care Hospital, China. All participants confirmed their participation by signing the consent forms. GDM diagnosis was conducted according to the World Health Organization (WHO) criteria.16 A total of 142 GDM subjects were included in case group, and 197 healthy subjects without any maternal gestational complication were included in control group. All participants have not smoking or drinking history during gestation. Meconium samples of newborns were collected in the first and two postnatal days, and urine samples were collected within 24 h after birth using newborn diapers according to our previously reported protocol.15,17 339 meconium samples (197 controls and 142 cases) and 177 responding urinary samples (96 controls and 81 cases) were included in this study. All the samples were stored at -80oC until further treatments.

Sample Preparation

Meconium was removed from the diaper, dried by a freeze dryer (Boyikang Corporation, Beijing, China) and ground into powder. 5 mg (dry weight) meconium was accurately weighed in a 1.5 mL centrifuge tube, and extracted according to the previously described method with some modifications.18 Different extraction solvents (i.e. methanol, acetonitrile, methanol-chloroform (3:1, v/v), acetonitrile-chloroform (3:1, v/v), and water were tested. Methanol is observed to be a simple optimal solvent for maximizing extracts of meconium metabolome. Therefore, 1 mL methanol was added and vortexed strongly for 1 min. The mixture was subsequently sonicated for 30 min in



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ultrasonic water bath. Then it was centrifuged at 12,000 rpm, 4 °C for 10 min and the supernatant was removed and collected. These steps were repeated once more with the residue of the previous extraction. The supernatant from these two extractions was mixed together, dried using a Speedvac concentrator (Thermo Scientific Corp., USA.) and reconstituted with 300 µL 50% (v/v) methanol. The reconstituted samples were centrifuged at 12,000 rpm, 4 °C for 15 min, and the supernatant was transferred into sample vials ready for metabolomics study. Urine was released from the diaper according to the method described in our previous report.17 Briefly, calcium chloride (CaCl2, 97%, Acros Organics, New Jersey, USA) was added to the diaper gel absorbent and the urine was consequently released from the diaper gel and collected after centrifuged at 1,000 rpm for 10 min using the self-designed urine separating device. The obtained newborn urine samples were mixed 1:1 (v/v) with deionized water. The effect of diaper absorption and urine extraction was investigated by comparing the metabolic profile of a directly collected fresh adult urine sample with that of the same urine sample absorbed and then released from a diaper. It was confirmed that metabolome of urine released from diaper was not affected by the diaper collection and extraction process.

Metabolome Analysis

Meconium metabolic profile (Figure S1A in Supporting Information) was acquired using a Waters ACQUITY ultra performance liquid chromatography (UPLC) coupled to a Waters Q-TOF Premier mass spectrometer (Waters Corp., Milford, MA, USA). The chromatographic separations were performed on an AcclaimTM RSLC 120 C18 column (2.2µm, 120Ǻ, 2.1×100mm) (Thermo Scientific, USA). The mobile-phase A was water containing 0.1% formic acid, while mobile-phase B was


Page 6 of 41

Page 7 of 41

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


acetonitrile containing 0.1% formic acid. The linear gradient increased from 5% to 10% B in 1 min, and from 10% to 60% B during 10 min, from 60% to 80% B during 5 min, and increased to 100% B during 4 min then decreased to 5% B in 0.1 min, and then it kept for 2 min for equilibrium. The flow rate was 0.4 mL/min. Injection volume was 5 µL. The mass spectrometer was operated in V mode and in negative ion mode. The capillary voltage and cone voltage were set at 2.7 KV and 35 V, respectively. The source and desolvation temperature was set at 150 °C and 350 °C, respectively. The flows of cone gas and desolvation gas were set at 50 and 500 L/h. Leucine enkephalin was introduced into electrospray source as a lock mass. The data acquisition rate was set to 0.1 s and the scan range was 100-1000 m/z. Data was collected in centroid mode.

Urine metabolic profile (Figure S1B) was acquired using a Waters ACQUITY UPLC system (Waters, Milford, MA, USA) coupled to a Q Exactive mass spectrometer (Thermo Fisher, San Jose, CA, USA). Chromatographic separation was performed on an ACQUITY UPLC BEH C18 column (1.7µm, 100mm×2.1mm ID) (Waters, Milford, MA, USA). For each sample, the run time was 22 min at a flow rate of 0.35 mL/min. The mobile phases were methanol with 0.1% formic acid (A) and H2O with 0.1% formic acid (B). The programmed gradient was 0 min, 0% A; 9 min, 45% A; 15min, 100% A; 18 min, 100% A; 18.10 min, 0% A; 22 min, 0% A. The column was maintained at 35 °C and the injection volume was 5 µL. The mass spectrometer was operated with a heated electrospray ionization (HESI) source in negative ion mode with a scan range of 100 to 1000 m/z, 70000 resolution. Spray voltage was set at 2500 V. Probe heater temperature was set at 425 °C. Capillary temperature was set at 262.5 °C. Sheath gas was set at 12.50 arbitrary units. Nitrogen was used as carrier gas. Sheath gas flow rate, aux gas and sweep gas flow rate was set at 50, 13, and 1 L/min



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

respectively. The injection valve of mass spectrometer was switched to waste to remove salt (esp. CaCl2) in acquired urine samples before 1 min. Data was collected from 1-21min in centroid mode.

All the MS/MS experiments were carried out using the Q Exactive mass spectrometer (Thermo Fisher, San Jose, CA, USA) for further identification of potential biomarkers. MS/MS was performed by normalized collision energy (NCE) technology with 30% NCE for the biomarkers with resolution of 17500.

Quality control and quality assurance (QC/QA)

Strict quality control and quality assurance (QC/QA) was conducted in this study. The stability and reproducibility of chromatographic separation and mass measurement during the entire sequence were assessed.19 All samples were run in a randomized fashion to eliminate the complications due to the artifacts related to injection order and gradual changes of instrument sensitivity in the entire sequence. An aliquot of each meconium sample was taken and mixed to give a meconium quality control (QC) sample, which featured the entire meconium sample set. A urinary QC sample was prepared in the same way to feature the entire urine sample set. The QC samples were injected at regular intervals (every 20 samples) during the sequence. The relative intensities of randomly selected variables exhibited small coefficients of variation across QC samples (Figure S2 in Supporting Information). The tight clustering of in-run QC samples in PCA scores plot (Figure 1A and C) also demonstrated the good overall repeatability during the whole sequence, which indicated the data set was worthy for the following analysis.

Data Processing and Biomarker Identification


Page 8 of 41

Page 9 of 41

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Raw data were converted to Xcalibur files and processed with SIEVE software (version 2.1, Thermo Fisher, USA), which allows alignment and frame to give a table of accurate mass and retention time paired with the associated intensities for all the detected peaks. The main parameters for meconium samples were set as follows: frame time width = 0.2 min, m/z width (ppm) = 100, m/z max = 1000, m/z min = 100, retention time = 0.4-19 min. The main parameters for urine samples were set as follows: frame time width = 0.1 min, m/z width (ppm) = 10; m/z max = 1000, m/z min = 100, retention time = 1-16 min.

The procedures of data processing, biomarker screening and identification were performed with the method previously described.20 Briefly, the intensity of extracted variables was normalized to the total intensity to reduce the variations from sample injection and enrichment factor. Extracted variables shown in 80% samples or above in each group were kept, otherwise they were removed. The missing values were substituted by 1/2 minimum of the intensity detected for corresponding variable. The processed table was then Pareto-scaled and exported to SIMCA-P (version 11.5, Umetrics, Sweden) for multivariate analysis. Principal component analysis (PCA) was performed to reveal any clustering in an unsupervised manner. Projection to partial least-squares discriminant analysis (PLS-DA) was then used to improve group classification and screen potential biomarkers. Portion (80%) of the samples (113 cases and 157 controls for meconium, 63 cases and 76 controls for urine) was used as a training set to construct the PLS-DA model. The remaining 20% samples (29 cases and 40 controls for meconium, 18 cases and 20 controls for urine) were used as a test set to evaluate the model. Potential biomarkers were screened based on the significance of their contribution to group variation, which was quantified by the variable importance in the project (VIP). The permutation test (999 random permutations) was used to validate the PLS-DA model (Figure



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

S3). Strict criteria were adopted for biomarker screening: (1) VIP scores >2 (Figure S4), (2) jack-knifing confidence interval > 0, (3) the intensity difference of the variables between case and control group was significant with p < 0.01 for meconium and p < 0.05 for urine, respectively. The detailed method for biomarker identification was described in previous report.20, 21 Information on potential biomarkers was searched in the Human Metabolome Database (HMDB) (http://www.hmdb.ca/). According to the mass accuracy of mass spectrometers we used, the accepted mass differences were set at 20 mDa for meconium and 5 mDa for urine during biomarker identification. Candidate biomarkers were further validated by searching against reported structural information in the literatures.

Statistical Analysis

Statistical analysis was performed using SPSS 18 (SPSS Inc., Chicago, IL). Unless specified, otherwise, a p-value of less than 0.05 was considered statistical significance. T-test and non-parametric Mann-Whitney U test were applied to compare the differences between two groups, for normally distributed and non-normally distributed continuous variables, respectively. Fisher’s exact test was applied to compare difference for categorical variables with total observed frequency ≤ 40, otherwise, Pearson Chi-square test was applied. The dose-related trends were analyzed using binary logistic regression. The defined outcome of non-GDM or GDM was counted based on the quartile cutoffs of the biomarker abundances of the control group. The effects of the potential biomarkers on the disease outcome were expressed by adjusted odds ratios (AORs) with 6 GDM risk factors,15 i.e., maternal age, pre-pregnant BMI, gravidity, parity, hepatitis B virus (HBV) infection (HBsAg positive), and newborn sex.


Page 10 of 41

Page 11 of 41

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Receiver Operating Characteristic (ROC) Analysis

The ROC curve analysis was performed using SPSS 18 (SPSS Inc., Chicago, IL). The training set was used to establish ROC model. Area under ROC curve (AUC) was used as a metric of sensitivity and specificity of the biomarkers. By searching sensitivity and specificity, the best cut-off point was determined for each biomarker. The test set was then subjected to the ROC model to evaluate its diagnostic ability. Since the combination of biomarkers usually is more effective than single biomarker,20-22 multi-biomarker models were created using ROCCET (ROC Curve Explorer & Tester, http://www.roccet.ca).23

RESULTS

Participant Demographics

Demographic data of the participants and their newborns are summarized in Table 1. Significant differences were observed in maternal age, pre-pregnant BMI, gestational age, gravidity, parity, mode of delivery and HBV infection between GDM and control group.

Meconium and Urinary Metabolic Profiling

Metabolic differences were observed by comparing the chromatograms of a GDM case versus control (Figure S1). PCA was initially performed on the entire data set to cluster the samples (Figure 1A and 1C). This non-supervised PCA analysis produced a weak clustering of these groups. A supervised analysis, PLS-DA, was further conducted using the training set. The permutation test (999 random permutations) validated the model, and no over fitting of the data was observed (Figure S3).



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

The scores plot (Figure 1B and 1D) shows a distinct segregation, indicating significant metabolic disturbance in the GDM cases. Four exogenous (mainly originated from food intake) metabolites (i.e. methylepicatechin, methylxanthine, dimethyluric acid, vanilloylglycine)24 and ten endogenous metabolites were identified for meconium (Table 2). Three biomarkers were screened out for urine (Table 2). More details about these potential metabolic biomarkers were described in Table S1.

GDM Risk Assessment of the Biomarkers

The potential biomarkers were subjected to ROC analysis and binary logistic regression to assess discriminative ability of GDM risk. A prognostic model is good when AUC ranges from 0.9 to 1, moderate if AUC is 0.7 to 0.9, and poor if AUC is 0.5 to 0.7.20 The AUC values of all meconium biomarkers were between 0.7 and 0.9 (Table 2), indicating their moderate discrimination. The AUCs of five endogenous biomarkers from meconium (i.e. argininosuccinic acid, methyladenosine, tetrahydrodipicolinate, DHAP (8:0), methylguanosine) were >0.8 (Table 2). The discriminative ability of the endogenous meconium metabolic biomarkers was further tested by the test set. The risk assessment scale (Table S3) showed that both high true positive ratio and true negative ratio ( > 75%) were obtained for argininosuccinic acid, methyladenosine and methylguanosine.

Multi-biomarker models were established for endogenous meconium biomarkers by using ROCCET - a web based tool (Figure S5). The combinations of nine meconium biomarkers (Figure 2A) turns to be the best GDM indicator with the highest AUC = 0.946 (true positive ratio 75.86 % and true negative ratio 95% in test set). The predictive accuracy of this nine-meconium biomarker model was showed in Figure S6.


Page 12 of 41

Page 13 of 41

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


The association of endogenous meconium biomarkers with GDM risk was further evaluated by binary logistic regression (Figure 3A and 3B). Crude odds ratios (ORs) were showed in Figure S7A and S7B. More details were showed in Table S2. Five biomarkers (i.e., argininosuccinic acid, methyladenosine, tetrahydrodipicolinate, DHAP(8:0) and taurodeoxycholic acid) are positively correlated with AORs (positive biomarkers), whereas other biomarkers (i.e., methylguanosine, glycocholic acid, hydroxyindoleacetylglycine, oxotrihydroxyleukotriene B4 and uric acid) are negatively correlated with AORs (negative biomarkers). Comparing the first quartile (set AOR=1) of the positive biomarkers, the AORs for argininosuccinic acid, tetrahydrodipicolinate and DHAP (8:0) significantly increased from the third to fourth quartile; AORs for methyladenosine, and taurodeoxycholic acid showed a significant increase at fourth quartile. Comparing the forth quartile (set AOR=1) of the negative biomarkers, the AOR for all five biomarkers increased from third to first quartile, but their AORs were only significant at the first quartile.

With regard to urine biomarkers, their AUCs are around 0.6 (Table 2), and the risk assessment scale of test set revealed their weak discrimination (Table S3). The effects of these three urine biomarkers on the GDM were further evaluated by ORs (Figure S7C, Table S2) and AORs (Figure 3C, Table S2). For the positive biomarkers (set AOR=1 at first quartile), AOR of uric acid increased from second to fourth quartile and it was significant at the fourth quartile. AOR of uridine increased at third and fourth quartile but no significance was observed at any quartile. For negative biomarker estrone (set AOR=1 at fourth quartile), the AOR increased from third to first quartile but no significance was found, either. In addition, the three urinary biomarkers failed to give an effective multi-metabolite biomarker model by ROCCET.



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 14 of 41

DISCUSSION

Newborn meconium and urine analysis: promising tools

Meconium analysis has been used to indicate fetal exposure to a variety of xenobiotic agents, including drugs, alcohol metabolites and nicotine metabolites.25 Meconium has gained wide acceptance in the scientific and medical communities for advantages, such as provision of a longer historical record, non-invasive analysis, and easier sample collection.26 Meconium is a repository of many metabolites and the first set of stool of a newborn infant. Therefore, it provides a wider window to detect the deposition of metabolites during gestation. However, so far there is a lack of published literatures regarding the application of newborn meconium metabolomics in characterizing maternal diseases. In this study, we used newborn meconium as a promising tool to study historical metabolic alternations induced by maternal disease like GDM. Meanwhile we tried to explore the potential of newborn urine in identifying metabolic pattern affected by maternal condition. To the best of our knowledge, this is the first study to investigate whether newborn meconium and urinary metabolic signatures can be indicative of metabolic disturbance induced by GDM during pregnancy.

Our results show that the meconium metabolome could be used to discriminate GDM cases from healthy control subjects. The alterations of meconium metabolome are much more significant than urinary metabolome. Interestingly, meconium GDM samples were separated into two clusters in PLS-DA plot. The exact reason is unclear but we speculate that there may be two causes: (1) the nature of meconium is far more complicated than biological fluid,25,

26

and the composition of

meconium may be highly different for individuals; (2) some GDM cases accompanied with maternal


Page 15 of 41

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


gestational complications (e.g. hepatitis B virus infection, gestational hypertension). PLS-DA analysis we used is capable to eliminate the potential adverse effects of individual difference and the complications on multivariate models. To screen most significant meconium biomarkers, a stricter criterion (significant difference of variables with p < 0.01 between case and control group, instead of p < 0.05) was conducted. A total of fourteen potential metabolic biomarkers in meconium were tentatively identified, which were mainly involved in purine metabolism, lipid metabolism, amino acid metabolism, cocoa and tea metabolism.

Data from newborn urinary metabolome analysis did not yield a clear discrimination of GDM cases and controls. The present observation was corresponding to the previous reports on maternal urine9,10. We tentatively propose that urinary profile analysis may not be a prudent choice for the study of GDM. In this study, diluted urine samples were directly injected into liquid chromatography for profile acquisition, which followed the well-established protocol by Want et al.19 Recently, in gas chromatography/mass spectrometry based urine metabolomics studies, urease is used to reduce urea level and subsequent chromatographic interference. However, the effect of urease treatment remains controversial in LC/MS based studies 27. Perhaps the treatment may improve the urinary metabolome discrimination for GDM cases, but it needs to be confirmed in the future. Three biomarkers were screened out from newborn urinary metabolome (uric acid, uridine and estrone) and their AUC values (Table 2) are quite poor, indicating that the pattern of newborn urinary metabolites was not influenced strongly. In addition, urinary uric acid tended to increase in GDM case group while meconium uric acid tended to decrease, which may reflect the different matrix effects. There is no denying that newborn urine is not equal to fetal urine and it only reflects a transient time window. Postnatal milk intake might also affect the urinary metabolome. However, it is still a new attempt.



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

The urinary biomarkers were generally involved in nucleic acid metabolism and lipid metabolism (steroids), which is roughly consistent with the metabolic alterations reflected by meconium metabolome analysis. This further confirms that nucleic acid metabolism (esp. purines) and lipid metabolism dysfunctions may be main effects induced by maternal GDM.

Exogenous meconium biomarkers

Four meconium metabolic biomarkers, i.e. methylepicatechin, methylxanthine, dimethyluric acid and vanilloylglycine (Table 2) were exogenous metabolites which are commonly present in green teas, red wine, cocoa products, and many fruits.28 Elevated levels of methylepicatechin, methylxanthine and dimethyluric acid suggested the increased consumption of a diet rich in fruit and vegetables in GDM mothers. We speculate that the GDM mothers might adjust their dietary composition according to doctor’s suggestion, and more flavanols were included in the menu due to the ability to relieve diabetes.29

Endogenous meconium biomarkers

Alterations of endogenous metabolites are observed when physiological state balance is disrupted. In the present study, a total of ten endogenous meconium biomarkers were tentatively identified, which indicated the alterations of metabolic network during GDM. Lipid metabolism, amino acid metabolism and purine metabolism were significantly disrupted metabolic pathways.

Lipid metabolism: Changes in carbohydrate and lipid metabolism always occur during pregnancy to ensure a continuous supply of nutrients to the growing fetus despite intermittent maternal food intake. These metabolic changes are progressive and may be accentuated in GDM women because


Page 16 of 41

Page 17 of 41

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


insulin action in late normal pregnancy is 50-70% lower than in non-pregnant women but metabolic adaptations do not fully compensate in GDM and glucose intolerance ensues.30 Four meconium metabolic biomarkers (taurodeoxycholic acid, glycocholic acid, oxotrihydroxyleukotriene B4, and DHAP (8:0)) are tightly associated with lipid metabolism. Taurodeoxycholic acid is formed by conjugation of deoxycholate with taurine, and glycocholic acid is an acyl glycine and a bile acid-glycine conjugate. Bile acids modulate the absorption of dietary fats and vitamins, and have been implicated to be important signaling molecules in cholesterol homeostasis and glucose control.31,32 In our study, the alterations of meconium taurodeoxycholic acid and glycocholic acid levels may reflect the disturbance of bile acid pool in GDM case group. The altered fatty acid oxidation has been confirmed in diabetes33 and GDM also induces a state of dyslipidemia (e.g. lower oxidation of fatty acid and triacylglycerol, increased synthesis of very low density lipoprotein (VLDL)).30,34 Oxotrihydroxyleukotriene B4 is the metabolite of omega-oxidation of leukotriene B4 (LTB4, a metabolite of arachidonic acid). It decreased in GDM group, indicating an abnormal fatty acid metabolism. Another potential biomarker, DHAP (8:0), is a member of DHAPs (dihydroxyacetone phosphates) which link glycerol metabolism to glycolysis. Increase of DHAP (8:0) in GDM cases may illustrate the disorders of lipid and carbohydrate metabolism. In addition, alkyl-DHAPs might also intermediate the synthesis of ether phospholipids. Since bile acids and phospholipids help lipid transportation, lipoprotein synthesis and transportation,35 an increase of lipoproteins in GDM cases was speculated. Although information about the maternal levels of triacylglycerol or lipoproteins of the participants was not available in our study, the decreased oxotrihydroxyleukotriene B4 and increase of DHAP (8:0), as well as altered bile acid conjugates, suggested a suppression of fatty acid oxidation in GDM, which may allow greater availability of



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 18 of 41

triacylglycerol to the fetoplacental unit cross the placenta. That may further contribute to fetal macrosomia.36 Our findings generally conform to the previous report of Dudziket al9 that the most pronounced GDM-specific changes corresponded to the lipid pattern (phospholipids, taurine-conjugates bile acids and long-chain polyunsaturated fatty acid derivatives). Figure 4 shows the possible biological pathways

that

our

four

lipid

metabolism-related

biomarkers

involved

in.

Decreased

lysoglycerophospholipids were observed in Dudzik’s study9 were not found in our study. This may due to the difference of the biological samples between our study and Dudzik’s study. Our observations indicated a long-term accumulation, which could be a complementary to traditional plasma analysis.

Amino

acid

metabolism:

Argininosuccinic

acid,

tetrahydrodipicolinate

and

hydroxyindoleacetylglycine link to amino acid metabolic processes, suggesting that the amino acid profile of meconium was dramatically altered by GDM progress. Different plasma levels of amino acids7, 37 and altered placental amino acid exchange38 in GDM mothers have also been reported previously. Arginosuccinic acid is a precursor of arginine, and tetrahydrodipicolinate is converted from L-aspartate. Hydroxyindoleacetylglycine is involved in tryptophan metabolism. These amino acids (e.g. arginine, aspartate) are the intermediates of important pathways like urea cycle and citric acid cycle. GDM may play a role in regulating these cycles by altering amino acid levels. In addition, GDM has been reported to be associated with altered L-arginine transportation and NO synthesis.39 Increased level of arginosuccinic acid in GDM group may reflect that the L-arginine/NO signaling pathway was also disrupted. Together with previous findings derived from maternal plasma or


Page 19 of 41

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


placenta, our findings from newborn meconium suggested that maternal hyperglycaemia was characterized by increased levels of gluconeogenic substrates.7

Purine metabolism: Three purine alkaloid metabolites (i.e. methyladenosine, methylguanosine and uric acid) were tentatively identified, indicating the dysfunction of purine metabolism in GDM. Augmented purine degradation has been found in diabetes patients.40 Clinical significance of six related purine metabolites (adenosine, adenine, inosine, xanthine, hypoxanthine and uric acid) was assessed in diabetes patients’ plasma and the results showed that levels of adenosine, inosine, uric acid and xanthine may be useful for monitoring the progression of diabetes.41 Uric acid is the final oxidation product of purine metabolism, and its level is lower in GDM cases with relative to controls in this meconium metabolomics study. This observation conformed to a previous serum study which reported that uric acid level tended to increase in non-diabetic individuals, but decrease in diabetic individuals.42 Besides, GDM is associated with endothelial dysfunction.43 Adenosine not only modulates energy homeostasis (ATP metabolism)44 but also activates the endothelial L-arginine/NO signaling pathway.45 Moreover, phosphate forms of guanosine (e.g. cyclic guanosine monophosphate (cGMP)) could act as second messenger in NO signaling pathway.46 Despite that these signaling molecules were not directly monitored in our study, their derivatives (i.e. methyladenosine, methylguanosine and argininosuccinic acid) significantly changed in GDM group. Methyladenosine, methylguanosine and argininosuccinic acid were also the indicators of NO signaling changes and they further imply that GDM may disrupt purine metabolism through L-arginine/NO signaling pathway.

Discriminative power of endogenous meconium biomarkers



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Arginosuccinic acid showed the highest AUC among single biomarkers (Table 2) and relatively strong GDM discriminative power in test set (true positive ratio 79.31% and true negative ratio 82.50%) (Table S3), and modified nucleosides methyladenosine and methylguanosine also exhibited moderate discriminative power in test set (true positive and negative ratio 75-83%) (Table S3), indicating their potential to become GDM biomarkers. Other biomarkers failed to achieve both high positive rate and high true negative rate in test set; as a result, they are not sensitive or specific enough for indicating GDM. Among meconium biomarkers, uric acid has the lowest AUC and low specificity, indicating its weakness in discrimination of GDM, although it used to be reported as a biomarker of diabetes risk in blood samples.41,42 The nine-meconium biomarker combination model revealed a much better discriminative ability (AUC > 0.9) than any single biomarker (Figure 2A). Its predicted class probabilities for each sample (Figure 2B) suggested a great potential for the use of this combined biomarker pattern in the accurate detection of GDM-induced disorders with high sensitivity. In summary, the ROC analysis results illustrate that argininosuccinic acid, methyladenosine and methylguanosine may be promising single GDM metabolic biomarkers and the combination of nine-endogenous meconium biomarkers improves the discriminative power compared to any single one.

Limitations of this study

This study provides novel insight into GDM-specific metabolic profile and biomarkers during gestation but much remains to be improved. (1) Although potential adverse effects from individual meconium composition difference and other gestational complications (e.g. hepatitis B virus infection, gestational hypertension) on statistical model was reduced to minimum in PLS-DA


Page 20 of 41

Page 21 of 41

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


analysis in this study, some uncertainty might be still involved in biomarker screening (two clusters in meconium PLS-DA plot). Therefore, more strict metrics of participant selection should be adopted in the future research. (2) Although the sample number in our study is already much larger than previous studies, a larger case-control study group is still in need to confirm these preliminary findings of this study. (3) A multiplatform of ‘omics’ is expected to improve our understanding of GDM-induced dysfunctions from meconium metabolome. Further studies conducted by different technologies like NMR or gas chromatography coupled to mass spectrometry can be good complements to this present study.

CONCLUSIONS

It is a new attempt to characterize the metabolic changes underlying GDM and identify possible biomarkers using newborn meconium and urine. This work shows that newborn meconium may be a valuable novel tool for monitoring historical metabolic alteration accompany with prenatal disorders. Several exogenous and endogenous biomarkers were screened out. The altered levels of endogenous biomarkers show dose-dependent associations with maternal GDM risk and indicate dysfunctions of lipid metabolism, amino acid metabolism and purine metabolism. Unbalanced lipid pattern is suggested to be a GDM-specific feature. Those results may provide novel insight into mother-fetal mutual metabolic processes that are specifically altered in GDM. However, the task of integrating these preliminary findings to clinically useful tools poses tremendous challenges (e.g. technical limitations, database challenges, costs and quality controls). More efforts are needed to achieve practical clinical applications.



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

FIGURES

Figure 1. Metabolome score plots following PCA (A: meconium; C: urine) and PLS-DA (B: meconium; D: urine) ● GDM case; ■control; △QC. The performance characteristics of meconium PCA multivariate model from a descriptive and predictive point of view were: R2(X) = 0.715; Q2(Y) = 0.443 while of urinary PCA multivariate model were: R2(X) = 0.743; Q2(Y) = 0.524. The performance characteristics of meconium PLS-DA multivariate model from a descriptive and predictive point of view were: R2(X) =0.244; R2(Y) = 0.848; and Q2(Y) = 0.728 while of urinary PLS-DA multivariate model were: R2(X) = 0.428; R2(Y) = 0.712; and Q2(Y) = 0.243.

Figure 2. ROC curve of the nine-meconium biomarker combination model (A) and its probability view (B). (●: case in training set; ○: control in training set; ●: case in test set; ○: control in test set. The nine-meconium biomarker:

argininosuccinic acid, methyladenosine, methylguanosine,

aurodeoxycholic acid, glycocholic acid, hydroxyindoleacetylglycine, oxotrihydroxyleukotriene B4, tetrahydrodipicolinate, DHAP (8:0).)

Figure 3. Dose-dependent association of endogenous positive meconium (A), negative meconium (B) and urinary (C) metabolic biomarkers with adjusted odds ratio (AOR) of GDM. (AOR: OR adjusted by newborn sex, maternal age, maternal pre-pregnant BMI, gravidity, parity and hepatitis B virus infection (HBsAg positive)).

Figure 4. Overview of the metabolic connections among the lipid metabolism related biomarkers in GDM. (Red arrow: increased in GDM; Green arrow: decreased in GDM; Metabolites with yellow background: biomarkers found in our study; Metabolites in blue: biomarkers identified in GDM maternal plasma from study of Dudziket al9. DHAP: dihydroxyacetone phosphate; AA methylester:


Page 22 of 41

Page 23 of 41

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


arachidonic acid methylester; DHA methylester: docosahexaenoic acid methylester; LPC: lysophosphatidylcholines; LPE: lysophosphatidylethanolamines; LPA: lysophosphatidic acid; LPI: lysophosphatidylinositol; LPS: lysophosphatidylserine).



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48

Page 24 of 41

TABLES

Table 1. Clinical data of the mothers and their newborns of metabolomics study. meconium (n=339)

urine (n=177)

training set

(n=29)

(n=40)

1.41E-01

3.33 ± 0.38

3.27 ± 0.32

26.14 ± 2.80

4.14E-05**

27.72 ± 4.50

21.98 ± 3.01

20.51 ± 2.14

1.90E-04**

39.15 ± 0.78

39.55 ± 0.96

5.25E-04**

Male

59 (52.21%)

77 (49.04%)

Female

54 (47.79%)

80 (50.96%)

67 (59.29%)

111 (70.70%)

Maternal age (year)b

(n=113)

(n=157)

3.27 ± 0.38

3.21 ± 0.29

28.07 ± 3.88

training set

control

weight (kg)a

control

test set case

Newborn birth

case

p

case

control

(n= 63)

(n=76)

5.07E-01

3.30 ± 0.39

3.20 ± 0.32

27.05 ± 2.40

4.66E-01

28.42 ± 4.79

21.66 ± 2.67

21.00 ± 2.75

3.17E-01

39.24 ± 0.89

39.44 ± 0.76

4.50E-01

15 (51.72%)

20 (50.00%)

14 (48.28%)

20 (50.00%)

16 (55.17%)

27 (67.50%)

8 (27.59%)

6 (15.00%)

5 (17.24%)

7 (17.50%)

27 (93.10%)

38 (95.00%)

2 (6.90%)

2 (5.00%)

0 (0.00%)

0 (0.00%)

20 (68.97%)

38 (95.00%)

p

test set case

control

(n=18)

(n=20)

1.25E-01

3.32 ± 0.42

3.22 ± 0.36

6.50E-01

26.13 ± 2.51

3.83E-03**

28.56 ± 4.26

27.65 ± 2.58

4.28E-01

21.89 ± 3.03

20.73 ± 2.33

5.33E-02

21.73 ± 2.92

20.51 ± 2.22

1.53E-01

39.11 ± 0.72

39.42 ± 0.92

3.00E-02*

39.12 ± 0.91

39.51 ± 0.77

2.28E-01

32 (50.79%)

37 (48.68%)

31 (49.21%)

39 (51.32%)

33 (52.38%)

51 (67.11%)

16 (25.40%)

17 (22.37%)

14 (22.22%)

8 (10.53%)

p

p

Maternal pre-pregnant 2 c

BMI (kg/m )

Gestational age (week)d Newborn sexe

6.08E-01

8.88E-01

8.04E-01

9 (50.00%)

8 (40.00%)

9 (50.00%)

12 (60.00%)

11 (61.11%)

13 (65.00%)

7.45E-01

Maternal gravidityf 1 2

25 (22.12%)

33 (21.02%)

≥3

21 (18.58%)

13 (8.28%)

1

90 (79.65%)

141 (89.81%)

2

21 (18.58%)

16 (10.19%)

2 (1.77%)

0 (0.00%)

75 (66.37%)

144 (91.72%)

3.22E-02

*

4.22E-01

1.14E-01

4 (22.22%)

6 (30.00%)

3 (16.67%)

1 (5.00%)

16 (88.89%)

20 (100.00%)

2 (11.11%)

0 (0.00%)

0 (0.00%)

0 (0.00%)

13 (72.22%)

18 (90.00%)

6.31E-01

Maternal parityg

≥3

3.10E-02*

7.39E-01

47 (74.60%)

65 (85.53%)

14 (22.22%)

11 (14.47%)

2 (3.17%)

0 (0.00%)

39 (61.90%)

72 (94.74%)

1.31E-01

2.18E-01

h

Mode of delivery Eutocia

8.42E-07**

3.55E-03**

24 / 36

ACS Paragon Plus Environment

8.72E-06**

2.22E-01

Page 25 of 41

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48

C-section Aid delivery


32 (28.32%)

12 (7.64%)

9 (31.03%)

2 (5.00%)

21 (33.33%)

3 (3.95%)

5 (27.78%)

2 (10.00%)

6 (5.31%)

1 (0.64%)

0 (0.00%)

0 (0.00%)

3 (4.76%)

1 (1.32%)

0 (0.00%)

0 (0.00%)

6 (20.69%)

0 (0.00%)

9 (14.52%)

0 (0.00%)

3 (16.67%)

0 (0.00%)

23 (79.13%)

40 (100.00%)

53 (85.48%)

76 (100.00%)

15 (83.33%)

20 (100.00%)

Maternal HBV infection (HBsAg positive)

i

Positive

19 (16.96%)

0 (0.00%)

Negative

93 (83.04%)

157 (100.00%)

2.24E-08**

3.96E-03**

a-d

5.29E-04**

9.67E-02

Values are expressed as mean ± SD. For meconium samples, difference of maternal age, pre-pregnant BMI and gestational age in both sets, between cases and controls, was compared by non-parametric Mann-Whitney U test; while difference of newborn birth weight in both sets, between cases and controls, was compared by independent sample t-test. For urine samples, difference of maternal age in both sets, and pre-pregnant BMI and gestational age, in training set, between cases and controls, were obtained by non-parametric Mann-Whitney U test, difference of newborn birth weight in both sets, maternal age and pre-pregnant BMI in test set, between cases and controls, was compared by independent sample t-test. e-i For meconium samples, difference between cases and controls of newborn sex, maternal gravidity and parity, mode of delivery in both sets, was compared by Pearson Chi-square test, while difference of HBV infection (HBsAg positive) in both sets, was compared by Fisher’s exact test. For urine samples, difference between cases and controls of newborn sex, maternal gravidity and parity, mode of delivery in training set was compared by Pearson Chi-square test; others were obtained by Fisher’s exact test. a-i * indicates p < 0.05, ** indicates p < 0.01. E-01 represents ×10-1.

25 / 36

ACS Paragon Plus Environment


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48

Page 26 of 41

Table 2. Tentatively identified potential meconium and urinary metabolic biomarkers of GDM and results of ROC analysis. Name

VIP

a

P value

b

Fold change c

AUC (95% CI, lower-upper bound)d

Class

Pathway

Meconium metabolic biomarkers Argininosuccinic acid 4.03 4.39E-26 ** 1.92 0.88 (0.84-0.92) ** Tetrahydrodipicolinate 2.72 3.99E-22 ** amino acid derivative amino acid metabolism 3.20 0.84 (0.80-0.89) ** ** ** Hydroxyindoleacetylglycine 2.45 1.22E-13 0.54 0.76 (0.70-0.82) ** Methyladenosine 4.19 2.25E-24 1.75 0.86 (0.82-0.91) ** Methylguanosine 2.51 2.44E-18 ** purine alkaloid metabolites purine metabolism 0.26 0.81 (0.76-0.87) ** ** ** Uric acid 2.88 2.35E-04 0.73 0.63 (0.56-0.70) ** Taurodeoxycholic acid 2.89 1.55E-09 1.54 0.71 (0.65-0.78) ** bile acid derivative Glycocholic acid 2.50 4.50E-16 ** 0.43 0.79 (0.73-0.85) ** lipid metabolism DHAP(8:0) 2.22 1.44E-18 ** 3.02 0.81 (0.76-0.86) ** DHAP derivative Oxotrihydroxyleukotriene B4 2.11 1.93E-10 ** 0.59 0.73 (0.66-0.79) ** fatty acid derivative ** ** Methylepicatechin 5.60 3.34E-19 5.71 0.82 (0.76-0.88) polyphenol host metabolites ** Vanilloylglycine 2.71 1.27E-14 0.46 0.78 (0.72-0.83) ** cocoa and tea metabolism Methylxanthine 2.21 1.64E-24 ** 6.82 0.86 (0.82-0.91) ** purine alkaloid metabolites Dimethyluric acid 2.25 1.29E-06 ** 1.76 0.67 (0.61-0.74) ** Urinary metabolic biomarkers 1.30 0.64 (0.54-0.73) ** purine alkaloid metabolites purine metabolism Uric acid 4.98 5.66E-03** * * Uridine 2.54 4.63E-02 1.14 0.60 (0.50-0.69) nucleoside nucleic acid metabolism * * Estrone 2.31 2.23E-02 0.75 0.61 (0.52-0.71) estrogen lipid (steroids) metabolism a Obtained from PLS-DA model with a threshold of 2. b Difference of biomarkers’ intensity between case and control group was compared. P values were calculated with Mann-Whitney test. c Fold changes were expressed by mean peak intensity of biomarkers in cases / mean peak intensity of biomarkers in controls. Case/control > 1 indicates a relatively higher concentration of this marker present in GDM patients, whereas < 1 indicates a relatively lower concentration of the marker in GDM patients as compared to healthy controls. b, d * indicates p

Newborn Meconium and Urinary Metabolome ... - ACS Publications

Recommend Documents