Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2022 Feb 7.
Published in final edited form as: Metabolomics. 2021 Feb 7;17(2):23. doi: 10.1007/s11306-021-01773-8

Global metabolomic profiling reveals hepatic biosignatures that reflect the unique metabolic needs of late-term mother and fetus

Nipun Saini 1, Manjot Virdee 1, Kaylee K Helfrich 1,2, Sze Ting Cecilia Kwan 1, Susan M Smith 1,2,3
PMCID: PMC8543356  NIHMSID: NIHMS1718549  PMID: 33550560

Abstract

Objective.

Gestational disorders including preeclampsia, growth restriction and diabetes are characterized, in part, by altered metabolic interactions between mother and fetus. Understanding their functional relevance requires metabolic characterization under normotypic conditions.

Methods.

We performed untargeted metabolomics on livers of pregnant, late-term C57Bl/6J mice (N=9 dams) and their fetuses (pooling 4 fetuses/litter), using UPLC-MS/MS.

Results.

Multivariate analysis of 730 hepatic metabolites revealed that maternal and fetal metabolite profiles were highly compartmentalized, and were significantly more similar within fetuses (ρaverage =0.81), or within dams (ρaverage =0.79), than within each maternal-fetal dyad (ρaverage = −0.76), suggesting that fetal hepatic metabolism is under distinct and equally tight metabolic control compared with its respective dam. The metabolite profiles were consistent with known differences in maternal-fetal metabolism. The reduced fetal glucose reflected its limited capacity for gluconeogenesis and dependence upon maternal plasma glucose pools. The fetal decreases in essential amino acids and elevations in their alpha-keto acid carnitine conjugates reflects their importance as secondary fuel sources to meet fetal energy demands. Whereas, contrasting elevations in fetal serine, glycine, aspartate, and glutamate reflects their contributions to endogenous nucleotide synthesis and fetal growth. Finally, the elevated maternal hepatic lipids and glycerol were consistent with a catabolic state that spares glucose to meet competing maternal-fetal energy demands.

Conclusions.

The metabolite profile of the late-term mouse dam and fetus is consistent with prior, non-rodent analyses utilizing plasma and urine. These data position mouse as a suitable model for mechanistic investigation into how maternal-fetal metabolism adapts (or not) to gestational stressors.

Keywords: Pregnancy, hepatic, untargeted metabolomics, amino acids, nucleotides, gestational stress

1. Introduction

Pregnancy is characterized by adaptive changes in maternal energetics and metabolism, driven primarily by the liver, that support the growing fetus. As the pregnancy advances, failures in these adaptive shifts have significant, deleterious consequences for fetal and infant health outcomes (Herrera 2000; Lain & Catalano 2007; Zeng et. al. 2017), as seen, for example, in maternal diabetes or malnutrition (Abu-Saad & Fraser 2010; Hambidge & Krebs 2018; Wu et. al. 2004). Insights into both normal pregnancy and these maladaptive changes have emerged from analysis of maternal-fetal metabolite profiles. These offer mechanistic insight into underlying pathologies, and identify biomarkers to predict at-risk pregnancies, as best exemplified in the widespread adoption of newborn bloodspot screening programs (Boardman et. al. 2019; McBride et. al. 2019; Mizejewski et. al. 2013; Traglia et. al. 2018). However, until recently, technical limitations in biochemical characterization drove an emphasis upon specific metabolites or metabolite classes (Hoffman et. al. 2019; Freedman et. al. 2018), and could not capture the full array of potentially relevant metabolites, including those providing novel mechanistic insight.

Untargeted metabolomics provides a comprehensive platform to interrogate how metabolic pathways both drive and respond to disease states, information essential for the design of functional interventions (Bardanzellu & Fanos 2019; Parfieniuk et. al. 2018; Sandlers 2017; Shaffer et. al. 2017). Its application to pregnancy is growing rapidly, and sampling of maternal plasma, amniotic fluid, urine, and placenta have informed conditions including preeclampsia, intrauterine growth restriction, preterm birth, infection, gestational diabetes (Carter et. al. 2019; Handelman et. al. 2019; Kelly et. al. 2017; Leite et. al. 2019; Sander et. al. 2019). However, the identification and interpretation of candidate biomarkers requires a strong understanding of maternal-fetal metabolite profiles during normal pregnancy, as these provide a framework for comparison and mechanistic understanding. A few studies have begun this characterization. For example, maternal plasma documents decreases in select amino acids, PUFAs, and acylcarnitines, and increased TCA intermediates and ketones with advancing gestation (Lindsay et. al. 2015). Comparison of maternal plasma and amniotic fluid revealed a metabolic switch-like transition from 2nd to 3rd trimester in healthy pregnant women, although only a few metabolites present in both biofluids could be correlated (Orczyk-Pawilowicz et. al. 2016). Global metabolic profiling of placental tissue from the maternal and fetal sides identified significant maternal enrichment in several amino acids, choline, glycerophosphocholine and TCA-cycle intermediates (Walejko et. al. 2018). Only recently is this approach being applied in rodent models to investigate gestational disorders such as diabetes, preeclampsia, and prenatal exposures (Akimova 2017; Brown 2017; Stojanovska 2019).

To address this gap, we used untargeted metabolomics to define the hepatic metabolite profile in the healthy, late-term maternal-fetal dyad, using a C57Bl/6J mouse. We tested the hypothesis that untargeted metabolomics would reveal biosignatures that distinguish mother from fetus, generate insight into their respective metabolic needs and processes, and create a novel framework of metabolite relationships that will inform the identification and mechanistic analysis of models of gestational stress.

2. Methods

2.1. Animals and Diets

Five-week old C57Bl/6J female mice (Jackson Laboratories, Bar Harbor, ME) consumed the AIN-93G purified diet (TD.94045, Envigo Teklad, Madison WI; Reeves et. al. 1993; Supplementary Table 1) throughout the study period. At 8 age weeks, they were mated with C57Bl/6J males; the morning of vaginal plug detection was embryonic day 0.5 (E0.5). On E17.5, pregnant dams (N=9) were killed by isoflurane overdose, and the maternal liver was removed, rinsed, and flash frozen in liquid nitrogen for metabolomic analysis. For fetal liver collections, the uterine horn was removed such that fetus closest to the right ovary of the dam is always considered as fetus 1. The livers from fetuses 1 to 4 were removed quickly, pooled and flash frozen for metabolomic analysis. Thus, maternal liver samples are 9 biological replicates, and fetal liver samples are the corresponding pooled fetuses (Fetus 1 to 4) from these 9 dams. The overall female-male sex ratio of the analyzed fetuses was 44:56. Samples were weighed at collection. All protocols were approved by the Institutional Animal Care and Use Committee of the David H. Murdoch Research Institute.

2.2. Metabolite Analysis

A global metabolomics approach was performed using a proprietary pipeline developed by Metabolon Inc. (Durham, NC). Samples were transported to Metabolon on dry ice and immediately stored at −80°C until processing. Samples were rapidly thawed, methanol containing several recovery standards was added proportionate to tissue weight, and the samples dissociated under vigorous shaking (2min, GenoGrinder 2000, Glen Mills) to denature the protein and release protein-bound small molecules. Protein concentrations were determined using the Bradford method for later data normalization. Samples were centrifuged to remove the protein precipitate, and the methanol extracts were harvested using an automated MicroLab STAR system (Hamilton Company, Salt Lake City, UT). Each sample was divided into five aliquots for their subsequent respective analyses, placed briefly on a TurboVap (Zymark) to remove the organic solvent, then stored overnight under nitrogen. Samples were reconstituted in the appropriate solvent, optimized for that analytical mode, and containing standards at fixed concentrations to normalize injection volume and chromatographic consistency.

Liquid chromatography-mass spectrometry (LC-MS) analysis was carried out using a Waters ACQUITY ultra high-performance liquid chromatography (UHPLC) and a Thermo Scientific Q-Exactive orbitrap mass spectrometer interfaced with a heated electrospray ionization source and Orbitrap mass analyzer operated at 35,000 mass resolution, as detailed in Ford (2020). Two samples were subjected to reverse phase (RP)/UPLC-MS/MS methods with positive ion mode electrospray ionization (ESI). In the first, which targeted more hydrophilic compounds, the extract was gradient eluted from a C18 column (Waters UPLC BEH C18–2.1×100mm, 1.7um) using water and methanol, and containing 0.05% perfluoropentanoic acid (PFPA) and 0.1% formic acid. The second positive ion mode ESI targeted more hydrophobic compounds, using the same C18 column with 0.05% PFPA and 0.1% formic acid in water, but gradient eluted using 50% methanol and 50% acetonitrile. A third aliquot was analyzed using RP/UPLC-MS/MS with negative ion mode ESI, using the same C18 column with 6.5mM ammonium bicarbonate at pH 8 in water, and gradient eluted using 95% methanol/5% water with 6.5mM ammonium bicarbonate at pH 8. The fourth sample was analyzed using HILIC/UPLC-MS/MS with negative ion mode ESI, using a Waters HILIC column (Waters UPLC BEH amide 2.1×150mm, 1.7um) with 15% water/5% methanol/80% acetonitrile against a gradient of 50/50 water and acetonitrile, all containing 10mM ammonium formate, pH 10.8. The fifth sample was reserved for backup. The separation and run times were ∼7min for the NEG and HILIC, and ∼3.5min for the POSEarly and POSLate, and used an alternating column system wherein one column performed the separation while the other was being cleaned and reconditioned for the next sample. The MS analysis alternated between MS and data-dependent MS scans using dynamic exclusion. The scan range varied slightly between methods but covered 70–1000 m/z.

Quality controls included a process blank of ultrapure water, a solvent blank, and a pooled matrix sample comprised of a small aliquot from each experimental sample. Five QC samples and three process blank samples were processed for every batch of 30 samples. Added to each experimental sample was a recovery sample cocktail of isotopically labeled and halogenated compounds selected to not interfere with measurement of endogenous samples, and used to aid chromatographic alignment and monitor instrument performance, as detailed (Ford 2020). As an additional quality control, the overall process variability was determined by calculating the median relative standard deviation (RSD) value for all endogenous metabolites present in 100% of the QC samples, which were technical replicates created from the pool of experimental samples. The median RSD for these QC samples was 8%. The injection order of the experimental samples was randomized, and the quality control samples were evenly interspersed between these injections. The samples analyzed here were all run in a single day. Additional details on the analytical methodology are presented in Ford (2020), which reports an average intra-assay coefficient of variation (CV) of <7%, and an average inter-assay CV of 9.9%−12.6%.

2.3. Data Analysis

Raw data were extracted, peak-identified, and the quality controls processed using custom software proprietary to Metabolon and based on criteria of peak detection, integration, and alignment (DeHaven 2010, 2012). Biochemicals were identified by comparison to Metabolon’s proprietary library, built from the analysis of authentic standards and comprised of 3300 purified, authenticated compounds annotated with respect to retention time/index (RI), mass-to-charge ratio (m/z), and chromatographic data including MS/MS spectral data on all four platforms (DeHaven 2010), consistent with Tier 1 identification standards defined by the Metabolomics Standards Initiative (Sumner 2007). Feature alignment across the samples was accomplished using a series of internal standards to establish a retention index (RI) ladder based upon the internal instrument performance standards. RIs of experimental peaks were determined by (i) comparison against the internal standard RI markers within 150 RI units (∼10 seconds) and assumed a linear fit, (ii) mass match to the library authenticated standard within +/− 10ppm, and (iii) quality of the fragmentation spectrum match between the experimental and library compound (Ford 2020). The use of all three criteria were used to distinguish and differentiate the biochemicals. All compounds reported were manually reviewed by a QC analyst to confirm the quality of peak integration, alignment, and identification across all samples in the study. Signals representing system artifacts, mis-assignments, and background noise were removed. Peaks were quantified using area-under-the-curve, and values are presented as median scaled data. All raw data provided by Metabolon are presented in Supplementary Table 2.

2.4. Statistical Analysis

We tested for unequal variance and normality within the two datasets using the Levene’s and Shapiro-Wilks tests, respectively, followed by analysis for significance using the Mann-Whitney U-test. Missing values were imputed using the minimum value obtained for that metabolite in that tissue. Any metabolite that was undetectable in at least five of nine samples was removed from the dataset. P-values were adjusted for multiple testing correction using the Benjamini-Hochberg False Discovery Rate (FDR; padj). Analyses were performed in the program R (version 3.6.1). Fold-change was defined as the metabolite’s abundance within fetal liver (FL) compared with maternal liver (ML).

For the discriminant analysis between maternal and fetal liver, the data were log transformed and auto-scaled (mean-centered and divided by the standard deviation of each variable), and then run through multivariate analysis (principal component analysis (PCA) and partial least square discriminant analysis (PLS-DA)), data visualization (heatmap and correlation matrix), and pathway analysis in MetaboAnalyst 4.0. PLS-DA variable importance projection (VIP) scores were used to identify each variable’s contribution to the model, using the leave-one-out cross-validation (LOOCV) method. VIP score was calculated as a weighted sum of the squared correlations between the PLS-DA components and the original variable, wherein the weights corresponded to the percentage variation explained by the PLS-DA component in the model. The number of terms in the sum depended on the number of PLS-DA components found to be significant in distinguishing the classes. To identify top pathways in each cluster, MetaboAnalyst 4.0 pathway analysis tool was used with KEGG IDs as input, selecting the following parameters for analysis: Mus musculus library (version Oct2019) as pathway library, “Globaltest” for pathway enrichment, and “relative betweenness centrality” as a measure for topological analysis. The boxplots were generated using ggplot2 package in R with maternal metabolite abundance set to 1.

3. Results

3.1. Metabolite profiles of maternal and fetal liver are distinct.

Untargeted metabolite analysis identified 854 metabolites in maternal and fetal liver. Of these, 56 were unique to either maternal or fetal samples, and 68 were detected in fewer than five samples and were excluded from further analysis. Of the remaining 730 metabolites, the most numerous were lipid-related (46.6%) and amino acid-related (19.5%); 66 had an unknown chemical structure. Supplementary Table 2 presents the identities of these metabolites, their fold-change in relative abundance (fetal abundance/maternal abundance), and adjusted p-values. The majority of these (84.9%) had significantly altered abundance within the fetal liver (FL) as compared with maternal liver (ML), 57.5% (N=343) with reduced abundance and 42.5% (N=254) with increased abundance in FL. Consideration of a P-adjusted between 0.05–0.10 flagged just an additional 3.2% of metabolites (N=23; N=13 increased, N=10 decreased). The metabolites having altered abundance spanned the range of structural/biochemical classifications including amino acids (82.4% out of 142), lipids (85.0% out of 340), nucleotides (91.8% out of 61), and carbohydrates (91.3% out of 46) (Supplementary Fig. 1). The tight variance for those metabolites having a differential abundance endorses observations that fetal liver metabolism is under tight regulatory control, and is distinct from that for maternal liver (Gruppuso & Sanders 2016; Rao et. al. 2013).

In further support of this hypothesis, multivariate analysis clearly separated the ML and FL metabolite profiles. In the PCA (Fig. 1a), PC1 defined 62.8% of the variance between these groups and PC2 contributed another 6.5%. PLS-DA provided similar results (Fig. 1b), and PC1 accounted for 62.8% of the variance and PC2 accounted for 5.3% of the variance (R2=0.9938; Q2=0.9873). In the PLS-DA model, variable importance projection (VIP) scores analysis identified the top 50 metabolites contributing to the separation in PC1 (Supplementary Table 3), and the top 30 metabolites are presented in Fig. 1c. Of the top 50 metabolites, the most common were lipids (36%), including five sphingomyelin derivatives, all reduced in the fetus. In contrast, fetal liver had striking enrichment in the phospho-sugars sedoheptulose-7-phosphate (45.2-fold), glucose-6-phosphate (96.4-fold), fructose-6-phosphate (34.1-fold), and mannose-6-phosphate (149-fold), as well as citrate (23.6-fold). Several vitamins and cofactors were significantly enriched in maternal liver including α-tocopherol (2500-fold), NAD (8.3-fold), FAD (4.3-fold), and nicotinamide (3.0-fold).

Figure 1. Multivariate Analyses of Fetal and Maternal Metabolite Profiles.

Figure 1.

Open in a new tab

In both the PCA (a) and PLSDA (b), PC1 clearly separated the maternal and fetal hepatic metabolite profiles and accounted for 62.8% of their variance. (c) VIP scores plot of the top 30 metabolites that contribute to Component 1 in the PLSDA. The red and blue boxes on the right indicate whether the mean metabolite abundance is increased (red) or decreased (blue) in fetal liver (FL) vs. maternal liver (ML). N=9 each for the maternal and fetal samples, N=730 for the metabolites.

Further informing the maternal-fetal hepatic differences were an additional fifty-six metabolites uniquely detected in either maternal (N=34) or fetal liver (N=22; Supplementary Table 4), and thus not captured in the VIP analysis. Complementing the reduced α-tocopherol, tocopherols and retinol were undetectable in fetal liver, even though liver is their primary storage site and endorsing that newborns obtain the bulk of their fat-soluble vitamins from lactation. The unique detection of six bile acids in maternal liver, along with a 125-fold enrichment in chenodeoxycholate, reflects their importance in bile acid synthesis and recycling, although some bile acids do cross the placenta (Wang et al. 2019). Vitamers related to folate and pantothenate were also undetectable in fetal liver. Most striking was the inverted relationship between the redox intermediates glutathione and cysteine, wherein the former and its metabolites were detected only in maternal liver, whereas cysteine and related metabolites were exclusive to fetal liver. Similar relationships are reported in human and mouse (Lambert et al. 1976; Raijmakers et al. 2001; Rollins et al. 1981), and may reflect the fetal need to prioritize cysteine for protein rather than glutathione synthesis. Supporting this was the unique fetal detection of five gamma-glutamyl-amino acid dipeptides, perhaps reflecting its three-fold enrichment in gamma-glutamyl transpeptidase activity (Rollins 1981). Also uniquely detected in fetal liver was sulfonated pregnanolone and its inactive metabolite pregnanediol.

3.2. Low interindividual variance in metabolite abundance within the respective fetal and maternal compartments

We used correlation analyses to investigate the within-group variance and overall trends in the metabolite profiles for ML and FL. Analysis using Euclidean distance and Ward’s linkages revealed that, consistent with the PCA and PLSDA, the ML and FL metabolite profiles were clearly differentiated (Fig. 2a). The majority of metabolites (left axis, Fig. 2a) had distinct, normalized abundance within each compartment (fetal vs. maternal), and the individual fetal metabolite profiles were more similar to each other, than they were to their respective dams. Few metabolites had a similar abundance between dam and fetus. This low, within-group variance was further endorsed in the correlation matrix (Fig. 2b), and the Spearman’s correlation values were largely positive between fetuses (ρ=0.65 to 0.91; p<0.0001) and between dams (ρ=0.56 to 0.89; p<0.0001) and negative against the fetuses’ respective dams or all the other dams (ρ= −0.63 to −0.84; p<0.0001) (Supplementary Table 5). Thus, maternal hepatic metabolites do not predict fetal hepatic processes.

Figure 2. Spearman’s Correlation Analyses of Metabolites in Individual Fetuses and Dams.

Figure 2.

Open in a new tab

a) Heatmap reveals that metabolite profiles within fetuses are more similar to each other, than they are to their respective dams. Each colored cell on the heatmap corresponds to a metabolite abundance value in the dataset, with individual fetal (F1–9) and maternal livers (M1–9) presented in columns and metabolites in rows. The high, unchanged to low abundance values of metabolites in maternal and fetal samples are represented by a gradient of red to green color. (b) Correlation matrix of samples again endorses that fetal liver metabolite profiles are more similar between individual fetal livers, compared with their dams. The strong positive correlations within fetuses and within dams are shown in red, whereas the negative correlations between fetuses and dams are shown in green. N=9 each for maternal and fetal samples, N=730 for the metabolites. (c) Correlation matrix of metabolites illustrate metabolite sets that are predictive of their respective hepatic compartments. A strong positive correlation between two metabolites is shown in red and negative correlation between two metabolites is shown in green color. No correlation between metabolites is shown in black color.

Given these high within-group correlations, we then asked if larger response patterns could be inferred within the collection of individual metabolites, irrespective of group identity. This revealed metabolite sets that shared a response within either ML or FL (green or red representing negative or positive correlation, respectively) (Fig. 2c). Smaller blocks of metabolites were not correlated (black) with other metabolites regardless of the compartment. These data further endorse that metabolic processes within fetus and dam are tightly regulated, and that these regulatory processes are distinct. It also affirms the existence of metabolite profiles that predict the normal functioning of their respective hepatic compartments.

3.3. Fetal hepatic metabolite profiles are consistent with pathways essential for its growth and development

We gained further insight into the maternal-fetal metabolite differences by mapping them using MetaboAnalyst (v4.0) (Supplementary Fig. 2). Pathways having the greatest differential FL/ML enrichment affected amino acid metabolism and included those for alanine/aspartate/glutamate (padj=5.91E-16), glycine/serine/threonine (padj=1.75E-14), and valine/leucine/isoleucine (padj=7.96E-14) (Table 1). Purine metabolism (padj=5.91E-16) was also a top altered pathway, as were energetic pathways involving citrate (padj=3.77E-16), pyruvate (padj=6.52E-15), thiamin (padj=1.28E-14), pentose phosphate (padj=1.75E-14), and nicotinate and nicotinamide (padj=3.77E-14).

Table 1.

Top 15 Pathways Enriched with Metabolites having Significantly Altered Abundance in Fetal liver, as Identified Using MetaboAnalyst 4.0.

Identified Pathways Total Compounds Hits Raw p −log P FDR Impact
Alanine, aspartate and glutamate metabolism 28 9 1.87E-17 16.73 5.91E-16 0.2508
Purine metabolism 66 20 1.91E-17 16.72 5.91E-16 0.2548
Citrate cycle 20 8 3.77E-16 15.42 6.52E-15 0.3590
Pyruvate metabolism 22 6 4.20E-16 15.38 6.52E-15 0.2493
Galactose metabolism 27 8 9.31E-16 15.03 1.15E-14 0.0733
Thiamine metabolism 7 4 1.24E-15 14.91 1.28E-14 0.6667
Glycine, Serine and Threonine metabolism 34 11 2.11E-15 14.68 1.75E-14 0.6816
Pentose Phosphate pathway 22 9 2.26E-15 14.65 1.75E-14 0.4991
Nicotinate and nicotinamide metabolism 15 6 5.47E-15 14.26 3.77E-14 0.5671
Arachidonic acid metabolism 36 3 7.65E-15 14.12 4.55E-14 0.3329
Aminoacyl-tRNA biosynthesis 48 15 8.63E-15 14.06 4.55E-14 0.1667
Glycerophospholipid metabolism 36 8 8.80E-15 14.06 4.55E-14 0.2775
Valine, leucine and isoleucine degradation 40 4 1.67E-14 13.78 7.96E-14 0.0524
Valine, leucine and isoleucine biosynthesis 8 3 2.21E-14 13.66 9.79E-14 0.0000
Butanoate metabolism 15 4 3.04E-14 13.52 1.23E-13 0.0000
Open in a new tab

However, a limitation of this analyses was that less than 40% of these metabolites were annotated in KEGG. We therefore complemented that analysis with visual inspection of the metabolites and pathways having altered representation. A majority (15/23) of the essential (0.34 to 0.68-fold) and non-essential amino acids (0.52 to 0.79-fold) had reduced abundance in FL, with the exceptions of aspartate, citrulline, glutamate, threonine, and tyrosine (Fig. 3a). In contrast, arginine (5.65-fold), cysteine (2.42-fold) and glycine (1.30-fold) were significantly enriched in FL.

Figure 3. Relative Abundance of (a) Amino Acids and (b) Purine and Pyrimidine Metabolites in Fetal vs. Maternal Liver.

Figure 3.

Open in a new tab

Maternal abundance is normalized to 1.0, and comparisons used Welch’s T-test. Boxplots depict the mean and variance for each metabolite class; dots indicate outliers. * padj<0.05. Abbreviations: AMP, adenosine-monophosphate; GMP, guanine-monophosphate; CMP, cytosine-monophosphate; UMP, uridine-monophosphate. Note that TMP was not detected.

Nucleotide metabolism was the second most enriched pathway, and FL had higher levels (2.11 to 4.73-fold) of all nucleosides except adenosine, elevated guanine (2.56-fold) and cytosine (4.13-fold), and reduced uracil (0.23-fold; Fig. 3b). In contrast, the abundance of purine breakdown products including hypoxanthine, xanthine, allantoin and urate was significantly reduced (0.10 to 0.65-fold).

Although the glycolytic and tricarboxylic acid pathways did not have significantly altered representation, visual inspection revealed that many of their metabolites were enriched in FL, including glucose-6-phosphate (96.4-fold), fructose-6-phosphate (34.1-fold), fructose-1,6-bisphosphate (10.3-fold), pyruvate (4.0-fold), citrate (23.6-fold), and succinate (1.44-fold) (Fig. 4a). Glucose itself was significantly lower (0.25-fold), as were phosphoenolpyruvate (0.51-fold), 3-phosphoglycerate (0.50-fold), and fumarate (0.82-fold).

Figure 4. Relative Abundance of (a) Glycolytic and TCA Cycle Intermediates and (b) Lipid Classes in Fetal vs. Maternal Liver.

Figure 4.

Open in a new tab

Lipid class are: medium chain FAs (C6:0-C12:0), saturated FAs (C14:0-C22:0), monounsaturated FAs (C14:1-C22:1), polyunsaturated FAs (C14:2-C24:6), phosphotidylglycerol (C16:0, C18:0, C18:1 and C18:2 at sn1 or sn2 position), phosphotidylcholine (C16:0, C18:0, C18:1, C18:2 at sn1; C18:2, C18:3, 20:4 and C22:6 at sn2 position), phosphotidylethanolamine (C16:0, C18:0, C18:1 at sn1; C18:1, C18:2, C20:4, C22:6 at sn2 position), phosphotidylserine (C16 and C18 side chains). Maternal abundance is normalized to 1.0, and comparisons used Welch’s T-test. Boxplots depict the mean and variance for each metabolite; dots indicate outliers. * padj<0.05. Abbreviations: Alpha-KG, alpha-ketoglutarate; DHAP, dihydroxyacetone phosphate; FA, free fatty acids; Fructose-6-P, fructose-6-phosphate; Fructose-1,6-BP, fructose-1,6-bisphosphate; glucose-6-P, glucose-6-phosphate; MUFA, mono-unsaturated FA; PEP, phosphoenolpyruvate; PUFA, poly-unsaturated FA; SFA, saturated FA.

With respect to lipids, apart from the medium chain fatty acids, FL had significantly lower free fatty acid levels including saturated (0.70-fold), monounsaturated (0.59-fold), and polyunsaturated (0.18-fold) fatty acids (Fig. 4b). Among phospholipids, phosphatidylglycerols (0.38-fold), phosphatidylcholines (0.76-fold), and phosphatidylethanolamines (0.89-fold) were also reduced, whereas phosphatidylserines were elevated (1.67-fold) and phosphatidylinositol levels did not differ.

4. Discussion

Using untargeted metabolomics, we established the hepatic metabolite profile of the late-term maternal-fetal dyad, using a mouse model fed a purified, fixed-nutrient diet. To our knowledge, this is the first untargeted maternal-fetal comparison, and the dataset’s informational richness facilitates a fuller understanding of their metabolic differences. Our core finding is that the mouse fetus, like other mammals, has a distinctive hepatic metabolite profile that does not mirror its mother’s, and instead is far more similar to those profiles from fetuses of independent pregnancies. This endorses numerous demonstrations that fetal metabolite pools are not generated through a simple ‘flow-down’ from maternal pools, but are the product of dynamic processes and regulatory mechanisms that meet its distinctive metabolic demands. As liver is the central regulator of metabolism (Berg et al. 2002), deviation from these profiles can offer mechanistic insight into the adaptations – healthy and pathological – that accompany maternal-fetal responses to stressors. Similar distinctions in human maternal-fetal metabolite profiles are reported for comparison of maternal-newborn plasma (Sakamoto & Kubota 2004), urine (Perrone et. al. 2020), and placental tissue (Walejko et. al. 2018), and have informed the development of predictive biosignatures for pregnancies at risk for adverse outcomes such as growth restriction and premature birth (Kadakia et. al. 2019; Maitre et. al. 2016; Scholtens et. al. 2016). Demonstration that this low inter-individual variance also applies to mouse endorses the model’s relevance for preclinical mechanistic investigations of pregnancy-related disease.

As a pregnancy progresses, essential metabolic adaptations within maternal liver alter maternal fuel choice and meet the demands of the growing fetus. The last trimester is considered a catabolic state for the mother, and is characterized by an increased lipolysis that spares glucose for fetal use (Butte 2000; Herrera & Ortega-Senovilla 2010). This lipolysis also provides fatty acid fuel for maternal energy demands, and glycerol substrate for gluconeogenesis (Lain & Catalano 2007; Zeng et. al. 2017), and the elevated glycerol and free fatty acids in ML are consistent with this lipolytic shift. In contrast, the fetus is highly anabolic, and in late-term even accrues glycogen and triglycerides for after-birth utilization (Battaglia & Meschia 1978). Maternal-derived glucose accounts for ∼80% of fetal energy consumption and, unlike its mother, the fetus cannot upregulate gluconeogenesis to address this energy gap unless severely stressed (Goodner & Thompson 1967; Jones 1976; Rao et. al. 2013). Thus, fetal-maternal ratios of glucose and gluconeogenic intermediates may inform the severity of metabolic stress in the fetus.

Because maternal glucose cannot fully meet fetal energy demands, the fetus also relies on the oxidation of amino acids, rather than lipids (Hay 2006; Rao et. al. 2013), as reflected in its sharply lower free fatty acid levels. Fetal plasma amino acids values typically exceed those of its mother (Cetin 2001; Manta-Vogli et. al. 2020), reflecting its reliance upon active placental transport to obtain essential and neutral amino acids. That many of these amino acids instead show 20% to 30% reductions in FL is inconsistent with these plasma findings, but consistent with their utilization through oxidative pathways to meet fetal energy demands. Further endorsing this are the FL’s five- to twenty-fold enrichment in the alpha-keto-acid carnitine conjugates of branched chain amino acids, which enter the mitochondria for beta-oxidation. The relatively modest fold-differences in hepatic urea, ornithine, citrulline, and aspartate, and the six-fold increase in fetal arginine, suggest significant urea cycle activity, even though the late-term fetus is highly anabolic. These findings complement the high rates of urea production previously documented for human fetuses and in preclinical animal models (Baig et. al. 1992; Battaglia & Meschia 1978; Raiha & Suinkonen 1968). Thus, the metabolite profile of these fetal livers is consistent with the importance of amino acid oxidation for fetal energy generation, and their elevation would signal an increased reliance on amino acid energy when the maternal glucose supply becomes limiting.

In contrast, fetal pools of aspartate, glutamate, glutamine, and glycine are governed by placental-fetal interactions, rather than through direct transfer from maternal pools (Cetin 2001; Kalhan 2013; Manta-Vogli et. al. 2020). Glutamine and glutamate exist in an inter-organ loop, in which the placenta amidates fetal hepatic glutamate to form glutamine, which then returns to fetal liver as a nitrogen source for the synthesis of metabolites such as purines and pyrimidines (Battaglia 2000; Battaglia & Meschia 1978). Thus, hepatic glutamine/glutamate content likely reflects the high fetal demand for glutamine nitrogen, and a reduced glutamine/glutamate ratio may signal an inadequate fetal nitrogen supply. A similar fetoplacental cycle exists for glycine and serine, wherein the placenta demethylates maternal serine and exports the glycine product into the fetal circulation. Fetal liver converts this glycine back to serine via serine hydroxymethyltransferase and the glycine cleavage system, generating methyl groups for the one-carbon pool, and ammonia for urea disposal and nitrogenous use (Cetin 2001; Kalhan 2013; 2016). The fetal glycine enrichment reflects both this interconversion and the high demand for glycine in purine synthesis, whereas the lower serine reflects its multiple contributions to one-carbon formation and placenta reuptake to regenerate more glycine carbon and nitrogen for the fetus (Kalhan 2013; 2016). Thus, fetal glycine/serine ratios also reflect nitrogen and methyl donor availability, and its measure may similarly inform the fetal capacity to adapt to stressors affecting growth and energy generation, and the mechanistic pathways underlying those metabolic shifts.

The high fetal demand for select amino acids also reflects their importance in nucleoside synthesis. Although the purified diet used here does not contain appreciable quantities of nucleoside bases, isotopic studies find that these contribute minimally, if at all, to maternal and fetal purine and pyrimidine pools, and thus their dominant source is de novo synthesis and recapture through the salvage pathway (Boza et. al. 1996). Liver is a major site of purine nucleoside and pyrimidine synthesis for itself and for other organs (Levine et. al. 1974; Murray 1971). Our data endorse this, as the KEGG pathways for purine and pyrimidine synthesis were significantly over-represented in the FL metabolite pool, as were aspartate/glutamate and glycine/serine synthesis; glycine, aspartate, and glutamine, along with one-carbon donation, contribute the carbon and nitrogen for the synthesis of purine nucleosides and pyrimidines (Kalhan 2013). Accompanying these differences were striking fetal enrichments in pentose phosphate intermediates including ribose-5-phosphate (107-fold), sedoheptulose-7-phosphate (45-fold), and fructose-6-phosphate (34-fold). However, these still represent a fraction of fetal hepatic glucose utilization, due to the additional contributions from the purine salvage pathway (Jones & Rolph 1985), reflected here as fetal elevations in inosine, GMP, and AMP, and decreases in catabolic products including hypoxanthine, xanthine, urate, and allantoin. In contrast, the pyrimidines uridine and cytidine readily cross the placenta, and this transfer increases during late-term growth (Hayashi et. al. 1968; Levine et. al. 1974) such that maternal uridine may comprise as much as 40% to 60% of fetal uridine (Gurpide et. al. 1972). The two- to three-fold fetal elevations in cytidine and uridine may reflect, in part, this directional flow.

This study has several limitations, the most important being that it represents a single snapshot at E17.5 and thus does not capture the adaptive maternal-fetal changes as the pregnancy advances. Moreover, although liver is a primary driver of metabolic adaptation between mother and fetus, it does not capture the additional metabolic influences of i.e., placenta, white adipose tissue, or skeletal muscle. Because the fetal livers could not be perfused, we elected to also not perfuse maternal liver; thus, these hepatic data include the contribution from blood. The fetal samples represent a pool of four fetuses; thus, we cannot identify potential sex-specific metabolite differences (Stojanovska et. al. 2019), although our PCA does not reveal an influence of fetal sex in these pooled samples (Virdee et.al., in review). Although these dams were not fasted, tissues were collected at a fixed time, 7.0 ± 1.5 hours into the light (sleep) cycle, and the tight variance and consistent metabolite changes across the maternofetal samples indicates this aspect of the study design is not problematic. Finally, although many rodent studies use fixed-formula diets, these dams consumed a purified diet from three weeks before and throughout pregnancy to standardize the reproducibility dietary contributions.

In summary, this is the first study to use a mouse model to evaluate the hepatic metabolite profiles of the maternal-fetal dyad. Their profiles at late-term are consistent with the distinctive metabolic differences between the mother and her fetus, and they complement an existing literature that largely focuses on humans and larger preclinical models such as sheep and non-human primates; moreover, they expand that literature to include liver, a major regulator of plasma metabolite pools. These data establish a baseline metabolite profile in a popular mouse model, for its future application to investigate maternal-fetal metabolic responses to disease states, and stressors ranging from gestational diabetes to nutrient malnutrition to toxicant exposure.

Supplementary Material

1718549_Sup_Tab_1

Table S1. Composition of the AIN-93G Purified Diet (TD.94045) Consumed by Mice in the Study.

1718549_Sup_Tab_2

Table S2. List of Metabolites Detected in Maternal and Fetal Liver with their Identifiers, Relative Abundance Fold-change (Fetal/Maternal) and Adjusted p-values.

1718549_Sup_Tab_3

Table S3. VIP Scores of the Top 50 Metabolites that Distinguish Component 1 in the PLS-DA.

1718549_Sup_Tab_4

Table S4. Metabolites Exclusively Detected in Either Maternal Liver or Fetal Liver.

1718549_Sup_Tab_5

Table S5. Spearman’s Correlation Coefficient values (ρ) and P-values of Maternal-Fetal Correlation Analysis Presented in Figure 2b.

1718549_Sup_Fig_1. Figure S1. Percent Distribution of All Metabolites in Fetal Liver Relative to Maternal Liver, Clustered by Biochemical Classification.

The numbers on bars indicate the number of metabolites in fetal liver that are significantly increased (red), decreased (green) or unchanged (yellow) in comparison to maternal liver. For each category, the righthand table details the number of metabolites identified, their percentage with respect to total metabolites (out of 730) and the percentage having significantly altered abundance.

1718549_Sup_Fig_2. Figure S2. Pathway Analysis of Metabolites Having Significantly Altered Abundance in Fetal Liver versus Maternal Liver.

The y-axis arranges the pathways based on p-values, whereas the x-axis represent the pathway impact based on pathway topology analysis. The node color reflects p value (increasing importance from white to red); the node radius reflects the pathway impact value. Analysis performed using MetaboAnalyst 4.0.

Acknowledgments

Funding: Supported by NIH awards R01 AA011085 and R01 AA022999 (SMS), F32 AA027121 (STCK), T32 DK007686 (KKH) and the UNC-NRI.

Supported by NIH awards #R01 AA022999 (SMS), #R01 AA011085 (SMS), #F32 AA027121 (STCK), #T32 DK007686 (KKH), and the UNC-NRI.

Footnotes

Declarations: All authors declare no conflicts or competing interests.

Data Availability: All data, processed and raw, are available as manuscript tables/figures or supplementary information.

Publisher's Disclaimer: This Author Accepted Manuscript is a PDF file of an unedited peer-reviewed manuscript that has been accepted for publication but has not been copyedited or corrected. The official version of record that is published in the journal is kept up to date and so may therefore differ from this version.

References:

  1. Abu-Saad K. and Fraser D. (2010). Maternal Nutrition and Birth Outcomes. Epidemiologic Reviews. 32(1):5–25. 10.1093/epirev/mxq001 [DOI] [PubMed] [Google Scholar]
  2. Akimova D, Wlodarczyk BJ, Lin Y, Ross ME, Finnell RH, Chen Q. et al. (2017). Metabolite profiling of whole murine embryos reveals metabolic perturbations associated with maternal valproate-induced neural tube closure defects. Birth Defects Research. 109(2):106–119. doi: 10.1002/bdra.23583 [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Baig MMA, Habibullah CM, Swamy M, Hassan SI, Uz-Zaman T, Ayesha Q, et al. (1992). Studies on Urea Cycle Enzyme Levels in the Human Fetal Liver at Different Gestational Ages. Pediatric Research. 31:143–145. 10.1203/00006450-199202000-00010 [DOI] [PubMed] [Google Scholar]
  4. Bardanzellu F, and Fanos V. (2019) The choice of amniotic fluid in metabolomics for the monitoring of fetus health – update. Expert Review of Proteomics. 16(6):487–499. [DOI] [PubMed] [Google Scholar]
  5. Battaglia FC and Meschia G. (1978). Principal substrates of Fetal metabolism. Physiological Reviews. 58 (2): 499–527. [DOI] [PubMed] [Google Scholar]
  6. Battaglia FC (2000). Glutamine and glutamate exchange between the fetal liver and the placenta. The Journal of Nutrition. 130(4S Suppl):974S–7S. doi: 10.1093/jn/130.4.974S [DOI] [PubMed] [Google Scholar]
  7. Berg JM, Tymoczko JL, and Stryer L. (2002). Section 30.2, Each Organ Has a Unique Metabolic Profile. Biochemistry. 5th edition. New York: W H Freeman. Available from: https://www.ncbi.nlm.nih.gov/books/NBK22436/ [Google Scholar]
  8. Boardman FK, Hale R, and Young PJ (2019). Newborn screening for haemophilia: The views of families and adults living with haemophilia in the UK. Haemophilia. 25(2):276–282. doi: 10.1111/hae.13706 [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Boza JJ, Jahoor F, and Reeds PJ (1996). Ribonucleic acid nucleotides in maternal and fetal tissues derive almost exclusively from synthesis de novo in pregnant mice. The Journal of Nutrition. 126(7):1749–1758. doi: 10.1093/jn/126.7.1749 [DOI] [PubMed] [Google Scholar]
  10. Brown AG, Tulina NM, Barila GO, Hester MS, and Elovitz MA (2017). Exposure to intrauterine inflammation alters metabolomic profiles in the amniotic fluid, fetal and neonatal brain in the mouse. PloS one. 12(10):e0186656. 10.1371/journal.pone.0186656 [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Butte NF (2000). Carbohydrate and lipid metabolism in pregnancy: normal compared with gestational diabetes mellitus. The American Journal of Clinical Nutrition. 71(5 Suppl):1256S–61S. doi: 10.1093/ajcn/71.5.1256s [DOI] [PubMed] [Google Scholar]
  12. Carter RA, Pan K, Harville EW, McRitchie S,and Sumner S. (2019). Metabolomics to reveal biomarkers and pathways of preterm birth: a systematic review and epidemiologic perspective. Metabolomics. 15(9):124. doi: 10.1007/s11306-019-1587-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Cetin I. (2001). Amino acid interconversions in the fetal-placental unit: the animal model and human studies in vivo. Pediatric Research. 49(2):148–154. doi: 10.1203/00006450-200102000-00004 [DOI] [PubMed] [Google Scholar]
  14. DeHaven CD, Evans AM, Dai H, and Lawton KA (2010). Organization of GC/MS and LC/MS metabolomics data into chemical libraries. Journal of Chemical Information. 2:9–21. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. DeHaven CD, Evans A, Dai H, and Lawton KA (2012). Software techniques for enabling high-throughput analysis of metabolomics datasets. InTech Open; doi: 10.5772/31277. http://www.intechopen.com/books/metabolomics/software-techniques-for-enabling-highthroughput-analysis-on-metabolomic-datasets. [DOI] [Google Scholar]
  16. Ford L, Kennedy AD, Goodman KD, Pappan KL, Evans AM, Miller LAD, et al. (2020). Precision of a clinical metabolomics profiling platform for use in the identification of inborn errors of metabolism. Journal of Applied Laboratory Medicine. 5(2):342–356. doi: 10.1093/jalm/jfz026. [DOI] [PubMed] [Google Scholar]
  17. Freedman R, Hunter SK, and Hoffman MC (2018). Prenatal Primary Prevention of Mental Illness by Micronutrient Supplements in Pregnancy. American Journal of Psychiatry. 175(7):607–619. doi: 10.1176/appi.ajp.2018.17070836. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Goodner CJ and Thompson DJ (1967). Glucose metabolism in the fetus in utero: The effect of maternal fasting and glucose loading in the rat. Pediatric Research. 1:443–451. [DOI] [PubMed] [Google Scholar]
  19. Gruppuso PA, and Sanders JA (2016). Regulation of liver development: implications for liver biology across the lifespan. Journal of molecular endocrinology. 56(3), R115–R125. 10.1530/JME-15-0313 [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Gurpide E, Tseng J, Escarcena L, Fahning M, Gibson C, and Fehr P. (1972). Fetomaternal production and transfer of progesterone and uridine in sheep. American Journal of Obstetrics and Gynecology. 113: 21. [DOI] [PubMed] [Google Scholar]
  21. Hambidge KM, and Krebs NF (2018). Strategies for optimizing maternal nutrition to promote infant development. Reproductive Health. 15(Suppl 1):87. 10.1186/s12978-018-0534-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Handelman SK, Romero R, Tarca AL, Pacora P, Ingram B, Maymon E. et al. (2019). The plasma metabolome of women in early pregnancy differs from that of non-pregnant women. PLoS One. 14(11):e0224682. doi: 10.1371/journal.pone.0224682. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Hay WW Jr. (2006). Recent observations on the regulation of fetal metabolism by glucose. The Journal of physiology. 572(Pt 1):17–24. 10.1113/jphysiol.2006.105072 [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Hayashi TT, Shin DH, and Wiand S. (1968). Placental transfer of orotic acid, uridine, and UMP. I. Comparison of acid-soluble and acid-insoluble counts. American Journal of Obstetrics and Gynecology. 102: 1144. [DOI] [PubMed] [Google Scholar]
  25. Herrera E. (2000). Metabolic adaptations in pregnancy and their implications for the availability of substrates to the fetus. European Journal of Clinical Nutrition. 54 Suppl 1:S47–S51. doi: 10.1038/sj.ejcn.1600984 [DOI] [PubMed] [Google Scholar]
  26. Herrera E. and Ortega-Senovilla H. (2010) Maternal lipid metabolism during normal pregnancy and its implications to fetal development. Clinical Lipidology. 5(6):899–911. [Google Scholar]
  27. Hoffman MC, Hunter SK, D’Alessandro A, Noonan K, Wyrwa A, and Freedman R. (2019). Interaction of maternal choline levels and prenatal Marijuana’s effects on the offspring. Psychological Medicine. 1–11. doi: 10.1017/S003329171900179X. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Jones CT (1976). Fetal metabolism and fetal growth. Journal of Reproduction and Fertility. 47:189–201. [DOI] [PubMed] [Google Scholar]
  29. Jones CT, and Rolph TP (1985). Metabolism during fetal life: a functional assessment of metabolic development. Physiological Reviews. 65(2):357–430. doi: 10.1152/physrev.1985.65.2.357 [DOI] [PubMed] [Google Scholar]
  30. Kalhan SC (2013). One-carbon metabolism, fetal growth and long-term consequences. Nestle Nutrition Institute workshop series. 74:127–138. 10.1159/000348459 [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Kalhan SC (2016). One carbon metabolism in pregnancy: Impact on maternal, fetal and neonatal health. Molecular and Cellular Endocrinology. 435:48–60. doi: 10.1016/j.mce.2016.06.006 [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Kadakia R, Nodzenski M, Talbot O, Kuang A, Bain JR, Muehlbauer MJ, et al. (2019). Maternal metabolites during pregnancy are associated with newborn outcomes and hyperinsulinaemia across ancestries. Diabetologia. 62:473–484. 10.1007/s00125-018-4781-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Kelly RS, Giorgio RT, Chawes BL, Palacios NI, Gray KJ, Mirzakhani H. et al. (2017). Applications of metabolomics in the study and management of preeclampsia: a review of the literature. Metabolomics 13, 86. 10.1007/s11306-017-1225-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Lain KY, and Catalano PM (2007). Metabolic changes in pregnancy. Clinical Obstetrics and Gynecology. 50(4):938–948. doi: 10.1097/GRF.0b013e31815a5494 [DOI] [PubMed] [Google Scholar]
  35. Lambert GH, and Thorgeirsson SS 1976. Glutathione in the developing mouse liver. I. Development curve and depletion after acetaminophen treatment. Biochemical Pharmacology 25:1777–1781. [DOI] [PubMed] [Google Scholar]
  36. Leite DFB, Morillon AC, Melo Júnior EF, Souza RT, McCarthy FP, Khashan A, et al. (2019). Examining the predictive accuracy of metabolomics for small-for-gestational-age babies: a systematic review. BMJ Open. 9(8):e031238. doi: 10.1136/bmjopen-2019-031238 [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Levine RL, Hoogenraad NJ, and Kretchmer N. (1974). A review: biological and clinical aspects of pyrimidine metabolism. Pediatric Research. 8(7):724–734. doi: 10.1203/00006450-197407000-00008 [DOI] [PubMed] [Google Scholar]
  38. Lindsay KL, Hellmuth C, Uhl O, Buss C, Wadhwa PD, Koletzko B, et al. (2015) Longitudinal Metabolomic Profiling of Amino Acids and Lipids across Healthy Pregnancy. PLoS ONE. 10(12): e0145794. doi: 10.1371/journal.pone.0145794 [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Maitre L, Villanueva CM, Lewis MR, Ibarluzea J, Santa-Marina L, Vrijheid M. et al. (2016). Maternal urinary metabolic signatures of fetal growth and associated clinical and environmental factors in the INMA study. BMC Medicine. 14:177. 10.1186/s12916-016-0706-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Manta-Vogli PD, Schulpis KH, Dotsikas Y, Loukas, and Y.L. (2020). The significant role of amino acids during pregnancy: nutritional support. The Journal of Maternal-Fetal and Neonatal Medicine. 33(2):334–340. doi: 10.1080/14767058.2018.1489795 [DOI] [PubMed] [Google Scholar]
  41. McBride KL, Pluciniczak J, Rhyand T. and Bartholomew D. (2019). Phenylalanine and tyrosine measurements across gestation by tandem mass spectrometer on dried blood spot cards from normal pregnant women. Genetics in Medicine. 21:1821–1826. 10.1038/s41436-018-0407-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Mizejewski GJ, Lindau-Shepard B, and Pass KA (2013) Newborn screening for autism: in search of candidate biomarkers. Biomarkers in Medicine. 7(2):247–260.doi: 10.2217/bmm.12.108 [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Murray AW (1971). The biological significance of purine salvage. Annual Review of Biochemistry. 40:811–826. doi: 10.1146/annurev.bi.40.070171.004115. [DOI] [PubMed] [Google Scholar]
  44. Orczyk-Pawilowicz M, Jawien E, Deja S, Hirnle L, Zabek A, and Mlynarz P. (2016) Metabolomics of Human Amniotic Fluid and Maternal Plasma during Normal Pregnancy. PLoS ONE. 11(4): e0152740. 10.1371/journal.pone.0152740 [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Parfieniuk E, Zbucka-Kretowska M, Ciborowski M, Kretowski A, and Barbas C. (2018). Untargeted metabolomics: an overview of its usefulness and future potential in prenatal diagnosis. Expert Review of Proteomics. 15(10):809–816. doi: 10.1080/14789450.2018.1526678 [DOI] [PubMed] [Google Scholar]
  46. Perrone S, Laschi E, De Bernardo G, Giordano M, Vanacore F, Tassini M. et al. (2020). Newborn metabolomic profile mirrors that of mother in pregnancy. Medical Hypotheses. 137:109543. doi: 10.1016/j.mehy.2019.109543 [DOI] [PubMed] [Google Scholar]
  47. Raiha NCR, and Suinkonen J. (1968). Development of urea synthesizing enzymes in human liver. Acta Paediatrica Scandinavica 57:121–124 [DOI] [PubMed] [Google Scholar]
  48. Raijmakers MTM, Steegers EAP, and Peters HM (2001). Glutathione S-transferases and thiol concentrations in embryonic and early fetal tissues. Human Reproduction. 16(11) 2445–2450. [DOI] [PubMed] [Google Scholar]
  49. Rao PNS, Shashidhar A, and Ashok C. (2013). In utero fuel homeostasis: Lessons for a clinician. Indian journal of endocrinology and metabolism. 17(1), 60–68. 10.4103/2230-8210.107851 [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Reeves PG, Nielsen FH, and Fahey GC Jr. (1993). AIN-93 purified diets for laboratory rodents: final report of the American Institute of Nutrition ad hoc writing committee on the reformulation of the AIN-76A rodent diet. The Journal of Nutrition. 123(11):1939–1951. doi: 10.1093/jn/123.11.1939 [DOI] [PubMed] [Google Scholar]
  51. Rollins D, Larsson A, Steen B, Krishnaswamy K. Hagenfeldt L, Moldeus P, and Rane A. (1981). Glutathione and gamma-glutamyl cycle enzymes in human fetal liver. Journal of Pharmacology and Experimental Therapeutics. 217:697–7000. [PubMed] [Google Scholar]
  52. Sakamato M. and Kubota M. (2004). Plasma Fatty Acid Profiles in 37 Pairs of Maternal and Umbilical Cord Blood Samples. Environmental Health and Preventive Medicine. 9:67–69 (short communication) [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Sandlers Y. (2017). The future perspective: metabolomics in laboratory medicine for inborn errors of metabolism. Translational Research. 189:65–75 [DOI] [PubMed] [Google Scholar]
  54. Sander KN, Kim DH, Ortori CA, Warren AY, Anyanwagu UC, Hay DP, et al. (2019). Untargeted analysis of plasma samples from pre-eclamptic women reveals polar and apolar changes in the metabolome. Metabolomics. 15(12):157. doi: 10.1007/s11306-019-1600-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Scholtens DM, Bain JR, Reisetter AC, Muehlbauer MJ, Nodzenski M, Stevens RD, et al. (2016). Metabolic Networks and Metabolites Underlie Associations Between Maternal Glucose During Pregnancy and Newborn Size at Birth. Diabetes. 65(7):2039–2050.DOI: 10.2337/db15-1748 [DOI] [PMC free article] [PubMed] [Google Scholar]
  56. Shaffer M, Armstrong AJS, Phelan VV, Reisdorph N, and Lozupone CA (2017). Microbiome and metabolome data integration provides insight into health and disease. Translational Research. 189:51–64. doi: 10.1016/j.trsl.2017.07.001 [DOI] [PMC free article] [PubMed] [Google Scholar]
  57. Stojanovska V, Dijkstra DJ, Vogtmann R, Gellhaus A, Scherjon SA, and Plösch T. (2019). A double-hit pre-eclampsia model results in sex-specific growth restriction patterns. Disease Models & Mechanisms. 12:dmm035980. doi: 10.1242/dmm.035980. [DOI] [PMC free article] [PubMed] [Google Scholar]
  58. Sumner LW, Amberg A, Barrett D, Beale MH, Beger R, Daykin CA, et al. (2007). Proposed minimum reporting standards for chemical analysis Chemical Analysis Working Group (CAWG) Metabolomics Standards Initiative (MSI). Metabolomics 3:211–221. [DOI] [PMC free article] [PubMed] [Google Scholar]
  59. Traglia M, Croen LA, Jones KL, Heuer LS, Yolken R, Kharrazi M. et al. (2018). Cross-genetic determination of maternal and neonatal immune mediators during pregnancy. Genome Medicine. 10(1):67. doi: 10.1186/s13073-018-0576-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  60. Virdee MS, Saini N, Kay CD, Neilson AP, Kwan STC, Helfrich KK, Mooney SM and Smith SM (2020). An enriched biosignature of gut microbiota-dependent metabolites characterizes maternal plasma in a mouse model of fetal alcohol spectrum disorders. Scientific Reports (Under Revision). [DOI] [PMC free article] [PubMed] [Google Scholar]
  61. Walejko JM, Chelliah A, Keller-Wood M, Gregg A, and Edison AS (2018). Global Metabolomics of the Placenta Reveals Distinct Metabolic Profiles between Maternal and Fetal Placental Tissues Following Delivery in Non-Labored Women. Metabolites. 8(1):10. doi: 10.3390/metabo8010010 [DOI] [PMC free article] [PubMed] [Google Scholar]
  62. Wang P, Zhong H, Song Y, Yuan P, Li Y, Lin S, et al. (2019). Targeted metabolomics analysis of maternal-placental-fetal metabolism in pregnant swine reveals links in fetal bile acid homeostasis and sulfonation capacity. American Journal of Physiology: Gastrointestinal Liver Physiology. 317:G8–G16. [DOI] [PubMed] [Google Scholar]
  63. Wu G, Bazer FW, Cudd TA, Meininger CJ, and Spencer TE (2004). Maternal Nutrition and Fetal Development. The Journal of Nutrition. 134(9):2169–2172. 10.1093/jn/134.9.2169 [DOI] [PubMed] [Google Scholar]
  64. Zeng Z, Liu F, and Li S. (2017). Metabolic Adaptations in Pregnancy: A Review. Annals of Nutrition and Metabolism. 70(1):59–65. doi: 10.1159/000459633. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1718549_Sup_Tab_1

Table S1. Composition of the AIN-93G Purified Diet (TD.94045) Consumed by Mice in the Study.

NIHMS1718549-supplement-1718549_Sup_Tab_1.docx (17.5KB, docx)
1718549_Sup_Tab_2

Table S2. List of Metabolites Detected in Maternal and Fetal Liver with their Identifiers, Relative Abundance Fold-change (Fetal/Maternal) and Adjusted p-values.

NIHMS1718549-supplement-1718549_Sup_Tab_2.xlsx (256.2KB, xlsx)
1718549_Sup_Tab_3

Table S3. VIP Scores of the Top 50 Metabolites that Distinguish Component 1 in the PLS-DA.

NIHMS1718549-supplement-1718549_Sup_Tab_3.pdf (250.1KB, pdf)
1718549_Sup_Tab_4

Table S4. Metabolites Exclusively Detected in Either Maternal Liver or Fetal Liver.

NIHMS1718549-supplement-1718549_Sup_Tab_4.pdf (90.1KB, pdf)
1718549_Sup_Tab_5

Table S5. Spearman’s Correlation Coefficient values (ρ) and P-values of Maternal-Fetal Correlation Analysis Presented in Figure 2b.

NIHMS1718549-supplement-1718549_Sup_Tab_5.pdf (299KB, pdf)
1718549_Sup_Fig_1. Figure S1. Percent Distribution of All Metabolites in Fetal Liver Relative to Maternal Liver, Clustered by Biochemical Classification.

The numbers on bars indicate the number of metabolites in fetal liver that are significantly increased (red), decreased (green) or unchanged (yellow) in comparison to maternal liver. For each category, the righthand table details the number of metabolites identified, their percentage with respect to total metabolites (out of 730) and the percentage having significantly altered abundance.

NIHMS1718549-supplement-1718549_Sup_Fig_1.tif (2.7MB, tif)
1718549_Sup_Fig_2. Figure S2. Pathway Analysis of Metabolites Having Significantly Altered Abundance in Fetal Liver versus Maternal Liver.

The y-axis arranges the pathways based on p-values, whereas the x-axis represent the pathway impact based on pathway topology analysis. The node color reflects p value (increasing importance from white to red); the node radius reflects the pathway impact value. Analysis performed using MetaboAnalyst 4.0.

NIHMS1718549-supplement-1718549_Sup_Fig_2.tif (4.3MB, tif)

RESOURCES