Heritable IUGR and adult metabolic syndrome are reversible and associated with alterations in the metabolome following dietary supplementation of 1-carbon intermediates

Abstract

Metabolic syndrome (MetS), following intrauterine growth restriction (IUGR), is epigenetically heritable. Recently, we abrogated the F2 adult phenotype with essential nutrient supplementation (ENS) of intermediates along the 1-carbon pathway. With the use of the same grandparental uterine artery ligation model, we profiled the F2 serum metabolome at weaning [postnatal day (d)21; n = 76] and adulthood (d160; n = 12) to test if MetS is preceded by alterations in the metabolome. Indicative of developmentally programmed MetS, adult F2, formerly IUGR rats, were obese (621 vs. 461 g; P < 0.0001), dyslipidemic (133 vs. 67 mg/dl; P < 0.001), and glucose intolerant (26 vs. 15 mg/kg/min; P < 0.01). Unbiased gas chromatography-mass spectrometry (GC-MS) profiling revealed 34 peaks corresponding to 12 nonredundant metabolites and 9 unknowns to be changing at weaning [false discovery rate (FDR) < 0.05]. Markers of later-in-life MetS included citric acid, glucosamine, myoinositol, and proline (P < 0.03). Hierarchical clustering revealed grouping by IUGR lineage and supplementation at d21 and d160. Weanlings grouped distinctly for ENS and IUGR by partial least-squares discriminate analysis (PLS-DA; P < 0.01), whereas paternal and maternal IUGR (IUGR^pat/IUGR^mat, respectively) control-fed rats, destined for MetS, had a distinct metabolome at weaning (randomForest analysis; class error < 0.1) and adulthood (PLS-DA; P < 0.05). In sum, we have found that alterations in the metabolome accompany heritable IUGR, precede adult-onset MetS, and are partially amenable to dietary intervention.—Seferovic, M. D., Goodspeed, D. M., Chu, D. M., Krannich, L. A., Gonzalez-Rodriguez, P. J., Cox, J. E., Aagaard, K. M. Heritable IUGR and adult metabolic syndrome are reversible and associated with alterations in the metabolome following dietary supplementation of one-carbon intermediates.

The initial studies by Sir David Barker and colleagues (1, 2) retrospectively linked low birthweight neonates to later risk of cardiovascular disease, describing the importance of the gestational environment to adult health. Associated conditions have now been extended to include poor neurologic development (3), compromised reproductive health (4), renal nephropathy (5), and metabolic-associated diseases, such as vascular dysfunction (6, 7), type II diabetes, dyslipidemia (8, 9), obesity, and hypertension (2, 10, 11).

It is now widely recognized that the gestational environment is a key arbiter of long-term health, aptly termed the developmental origins of health and disease or commonly referred to as fetal programming. With an aberrant fetal to adult phenotype, fetal programming has been demonstrated to occur across multiple generations (12–17) and be accompanied by epigenomic alterations through histone modification or methylation of key gene regulatory regions of the offspring’s genome that can be maternally and paternally inherited (17–21) and span generations (14, 22). Implicit in these collective observations is the underlying notion that gestation is a key intervention point, during which, the timely recognition and effective treatment of gestational pathologies have the potential to ameliorate health over an individual's lifetime, as well as subsequent generations. Thus, there is an apparent need for discovery of markers and intervention strategies.

Uteroplacental insufficiency, resulting in IUGR, often results from vascular malformations or underdevelopment of the spiral arteries in the placenta that lead to restricted maternal/fetal exchange, subjecting the fetus to an altered gestational environment characterized by substrate deprivation, hypoxia, acidosis, and accompanying inflammation. These IUGR neonates have a significantly increased risk of perinatal complications and compromised health in adulthood with conditions, such as MetS. Although developmental regulatory changes in utero as a result of IUGR may contribute to adult disease (23, 24), previous studies have demonstrated that metastable, noncoded gene expression modifiers (i.e., alterations to nucleosome positioning, histone modifications, alterations in DNA methylation, and variance at metastable epialleles) are potential mechanisms of later onset health conditions (25–28). Furthermore, IUGR has been shown to affect DNA methylation greatly (29, 30), and it is thought that hypoxia is able to induce epigenetic modifications (31). For instance, specific alterations in DNA methylation following IUGR have been shown in rat models following bilateral uterine artery ligation (30, 32), as well as in human cohorts (33, 34).

Methylation changes are highly sensitive to levels of the epigenetic-modifying nutrients, methionine, B₁₂, folic acid, betaine, choline, zinc, and arginine, many of which are essential components of 1-carbon metabolism [reviewed in refs. (35, 36)]. Supplementation or deprivation of these nutrients has been shown in human population studies and animal models to induce DNA methylation changes and thus, potentiate disease (36–39). Recently, we demonstrated that after inducing multigenerationally inherited adult MetS with a well-established bilateral uterine artery ligation model of IUGR in grandparental (P1) dams, ENS ameliorated multigenerational adult MetS, reversing the phenotype (40). Specifically, F2 pups of IUGR^pat/IUGR^mat lineage exhibited hallmark MetS characteristics of dyslipidemia, increased adiposity, and insulin resistance, whereas the IUGR^pat/IUGR^mat lineage ENS-fed rats were indistinguishable from the Sham^pat/Sham^mat linage controls fed at d160.

As IUGR^pat/IUGR^mat lineage rats predictably developed MetS on a control diet, we reasoned that underlying alterations in small molecular-weight metabolites (collectively referred to as the metabolome) could be detected and would reliably predict later in life onset of adult MetS. We hypothesized that significant altered metabolomic profiles occurring among IUGR lineage pups would precede the phenotype (i.e., at the time of weaning and before dietary intake) and that gestational and lactational exposure to an ENS diet would normalize the IUGR^pat/IUGR^mat serum metabolome in the F2 generation. Furthermore, the identification of early markers or biochemical changes preceding MetS may provide an insight into a means of identifying predictors and pathways integral to modulating the development of MetS. In this study, therefore, we characterized serum metabolome changes associated with multigenerational IUGR and the subsequent development of MetS in the F2 generation. Specifically, with the use of an unbiased (naïve) GC-MS approach, we have interrogated significant differences in the metabolomic signature and specific metabolites at weaning (d21) and thus, before the introduction of the diet and long before development of adult MetS. Moreover, we further interrogated the persistence or variance of these early metabolic markers among the F2 generation in adulthood (d160). By sampling both interval time-points among the exposure groups, we sought to discriminate the influence of ENS dietary modifications and IUGR lineage on the adult metabolic profile and to identify potential biomarkers in the metabolome that precede dietary intake and the onset of clinically evident adult MetS. In sum, the data reveal a modifiable serum metabolome, which is influenced significantly by IUGR lineage and by supplementation of 1-carbon intermediates in the parental and F2 diet. Several metabolite markers predictive of MetS were identified.

MATERIALS AND METHODS

Multigenerational model of MetS following IUGR

A detailed characterization of our model and procedures, including the development of MetS and the effect of ENS supplementation on the phenotype, was published (40). All experiments were conducted with review and approval of the University of Utah Institutional Animal Care and Use Committee. In brief, bilateral uterine artery ligation was performed to produce the uteroplacental insufficiency IUGR model on pregnant Sprague-Dawley rats on gestational (embryonic) day (e)19. Sham surgeries were performed for controls. On e21, pups were born via cesarean delivery, and on d21, offspring were weaned onto a lifelong diet of control rat chow (Harlan Teklad 8640; Harlan Laboratories, Indianapolis, IN, USA) or an isocaloric rat chow supplemented with essential nutrients, the constituents of which are detailed in Table 1. Mating pairs were established on d80. The F2 generation was weaned onto the same diet of its maternal and paternal lineage. Serum was collected, according to the schedule (see description in Fig. 1A, B).

TABLE 1.

Constituents of the ENS diet

Supplement	Amount of chow diet (Harlan-Teklad)
Folic acid	15 mg/kg
Choline	15 mg/kg
B12	1.5 mg/kg
Betaine	15 g/kg
l-Methionine	11 g/kg
l-Arginine	50 g/kg
Zinc	150 mg/kg

See Goodspeed et al. (40) for further details.

Figure 1.

IUGR-induced MetS model description. A) Pregnant Sprague-Dawley rats underwent sham or bilateral uterine artery ligation surgery at e19. At birth, litters were culled to 8 pups. On d21, mice were assigned a lifelong control (Cont) or ENS diet. At d80, mice were paired to a matching IUGR or Sham mate of the same diet. The F2 offspring were continued onto the corresponding maternal/paternal diet at weaning on d21. Blood serum was collected from F2 offspring at d21 and d160. B) Metabolomic profiling experiments were carried out on the blood as indicated. C) IUGR induces adult onset metabolic disease, apparent at d160. F2 male (Ci) and female (Cii) d160 weights, deoxyglucose uptake, and triglycerides on control (gray boxes, sham lineage control diet) or ENS (blue boxes, sham lineage ENS diet) diet. Red boxes, IUGR lineage control diet; magenta boxes, IUGR lineage ENS diet. Differences from IUGR^mat/IUGR^pat control fed are indicated (aP < 0.05, bP < 0.01, cP < 0.001, dP < 0.0001, 1-way ANOVA, Dunnett’s post hoc test). DG6P, 2-deoxy-glucose-6-phosphate uptake.

To characterize MetS phenotype, serum lipid profiling and hyperinsulinemic-euglycemic clamp studies were measured and carried out as described previously (40). Rats were deprived of food for 12 hours, and triglyceride concentrations were obtained by clinical laboratory testing, performed by ARUP Laboratories (Salt Lake City, UT, USA) by use of quantitative enzymatic assays. For hyperinsulinemic-euglycemic clamp procedures, rats were first deprived of food for 6 hours. Hyperinsulinemia was initiated with an intravenous injection of insulin. Glucose was maintained at a steady state of ∼130 mg/dl, 90 minutes after injection, by infusing dextrose. Next, euglycemia was maintained for 45 minutes. Glucose infusion rates were calculated as the average glucose infusion rate during the steady-state period. [³H]2-Deoxyglucose was given as a bolus during the steady state for estimation of 2-deoxyglucose uptake, as described previously (41).

MS metabolome profiling

Analysis of the serum metabolome was undertaken similar to a previous description (42). Frozen serum samples, stored at −80°C, were thawed, and proteins were separated, similar to as described (43). First, purified stable isotope variants of amino acids were added as internal reference controls. Then, a 360 µl volume of precooled 90% methanol (−20°C) was added to 40 µl sample volumes of the individual tubes for a final concentration of 80% methanol. The samples were then incubated for 1 hour at −20°C, followed by centrifugation at 30,000 g for 10 minutes by use of a precooled rotor. The supernatant containing the extracted metabolites was aspirated and transferred to fresh disposable tubes and lyophilized by use of a cooled vacuum centrifuge. A derivatization protocol was performed to volatize highly stable compounds. In brief, samples were resuspended and then incubated at 60°C for 30 minutes in 50 µl pyridine with 20 mg/ml O-methylhydroxylamine hydrochloride. Tubes were sonicated for 5 minutes in a water bath, and 50 µl N-methyl-N-(trimethylsilyl)trifluoroacetamide/1% trimethychlorosilane (Pierce, Rockford, IL, USA) was added immediately. Samples were incubated once more at 37°C for 30 minutes. Retention standards were added, consisting of fatty acid methyl esters before analysis.

Samples were analyzed by use of an Agilent 6890 gas chromatograph (Agilent, Santa Clara, CA, USA), coupled to a Micromass GCT Premier (Waters, Milford, MA, USA). Samples were analyzed in random order by use of an autosampler that injected 1 µl in a 2:1 split ratio with a constant temperature of 220°C. Separation was obtained with a 30 m Rtx-5MS, preceded by an Integra-Guard column obtained from Restek (Bellefonte, PA, USA) and by use of helium at a flow rate of 1 ml/min. An initial 2-minute temperature step at 70°C was increased to 250°C (8°C/min rate of increase) and then increased subsequently to 330°C (20°C/min rate of increase), which was maintained for 2 minutes. The transfer line into the mass spectrometer was held at 250°C. Normal electron impact ionization conditions were used, and raw spectral data were obtained with MassLynx software 4.1 (Waters).

Quantitative analysis and statistics

To identify differences between the groups, chromatographic deconvolution and area-under-the-curve analysis was performed in MarkerLynx (Waters). The retention time and peak area were then exported, and a standard analysis was undertaken by use of MetaboAnalyst pipeline software (v2.0; Edmonton, AB, Canada, www.metaboanalyst.ca) (44). Data were filtered by variance (sd, median absolute deviation/median), followed by square-root transformation and range scaling of the data (45). Significant changes were determined subsequently by 1-way ANOVA, followed by Fisher’s least significant difference post hoc test. All significant changes, as determined by P < 0.05, are reported by use of heat maps, whereas tables of identified metabolites are filtered for FDR. A composite table report changes as relative Z score compared with Sham^mat/Sham^pat lineage control fed (46).

Hierarchical cluster was determined by Pearson’s distance measure and Ward’s clustering algorithm and indicated by dendrogram. Multivariate analysis was performed by PLS-DA. A leave-one-out cross-validation was used to determine the optimal number of components by predictive squared correlation coefficient and permutation analysis, whereby rats are randomized and PLS-DA performed, with the probability measured, as the number of times permutated yielded greater significance. Samples were plotted with a 95% confidence area between groups. randomForest was performed to determine overall group differences in the complex samples. The class error (400 grown trees) was plotted. Significantly changing metabolites or metabolites of interest were identified based on their characteristic retention time and mass-to-charge (m/z) ratio by use of an in-house database or the NIST MS Search 2.0 (National Institute of Standards and Technology, Gaithersburg, MD, USA; chemdata.nist.gov). Select samples were plotted by use of hive plots (HiveGraph 1.0; Wodak Lab, Toronto, ON, Canada; wodaklab.org/hivegraph/graph) (47).

RESULTS

To test whether changes induced by IUGR could result in multigenerational metabolomic alterations, we established and have now described our uterine artery ligation model (40). Pregnant dams (grandmaternal P1) underwent uterine artery ligation or sham surgery, and their normal (sham) or growth-restricted (ligated) pups were assigned a control or 1-carbon intermediate ENS diet at weaning (Fig. 1A). Parental generation (F1) pairs matching in diet allocation and IUGR^mat/IUGR^pat or Sham^mat/Sham^pat lineage were mated, and their progeny continued on the parental diet. From the second generation offspring (F2), serum was collected at weaning in 3 separate experiments (n = 76) or at maturity (d160; n = 12; Fig. 1B). The F2 rats of IUGR^mat/IUGR^pat lineage developed MetS in adulthood (Fig. 1C). Specifically, the IUGR^mat/IUGR^pat lineage adult male rats fed a control diet exhibit an ∼30% weight increase and an ∼40% increase in deoxyglucose uptake and triglycerides compared with Sham^mat/Sham^pat lineage control-fed males (Fig. 1Ci). Additionally, adult females had a 15% increase in weight and an ∼35% increase in deoxyglucose uptake and triglycerides (Fig. 1Cii). As described previously in detail, this heritable phenotype of MetS, composed of obesity, glucose intolerance, and dyslipidemia, was ameliorated with ENS supplementation of the parental and F2 generation (Fig. 1C) (40).

To characterize metabolomic changes associated with the development of heritable MetS, serum was collected at weaning (d21) before the onset of MetS and also in adulthood (d160) when the phenotype was apparent (Fig. 1). Serum metabolites from 4 independent experiments were quantified by use of an unbiased GC-MS analysis. Peaks were associated with their respective metabolites by match to standard databases. Peaks were quantified based on their relative intensity and assessed for significance of difference by ANOVA. All significant changes (P < 0.05) were plotted onto heat maps. A representative of 3 independent experiments is shown in Fig. 2A. There were broad metabolite pattern differences among the 4 cohorts; however, hierarchical clustering by Pearson’s revealed that the expression profiles separated by dietary group, with ENS as the greatest determinant of dissimilarity. Regardless of dietary group, IUGR^mat/IUGR^pat and Sham^mat/Sham^pat lineage groups separated distinctly, particularly for the control-fed group (Fig. 2A). A composite table of all changing metabolites identified from the 3 experiments at weaning and their relative changes compared with the Sham^mat/Sham^pat lineage control-fed group are indicated in Table 2.

Figure 2.

Relative change in serum metabolite expression between groups. A) Changes between groups are indicated by shading for a representative experiment at the time of weaning (d21; n = 24). B) Changes in adulthood (day-of-life d160) after IUGR^mat/IUGR^pat-Cont develops MetS phenotype (n = 12). Replicate rats are indicated on the y axis, with metabolites on the x axis. Unidentified metabolites are blank. All metabolites, altered significantly in expression by 2-way ANOVA (P < 0.05), are included. Ward’s clustering algorithm was used with Pearson’s distance measure.

TABLE 2.

Metabolites and small molecule intermediates that significantly vary by virtue of heritable lineage and dietary exposure in utero and during lactation

Metabolite	Identification	Shammat/Shampat ENS	IUGRmat/IUGRpat Cont	IUGRmat/IUGRpat ENS	P	FDR
Unkown1	—	−0.06	−0.84	−0.13	<0.0001	<0.001
Glycine	HMDB00123	−0.73	−0.88	−0.65	<0.0001	<0.001
Unkown2	—	0.3	−0.48	0.29	<0.0001	0.001
Myoinositol	HMDB00211	−0.12	−0.77	−0.27	<0.001	0.001
Unknown3	—	0.34	−0.08	0.29	<0.0001	0.003
l-Methionine	HMDB00696	0.57	0.05	0.58	<0.0001	0.003
Citric acid	HMDB00094	0.43	−0.47	0.45	<0.0001	0.004
l-Methionine	HMDB00696	0.57	0.05	0.58	<0.0001	0.005
4-Hydroxyproline	HMDB00725	−0.5	−0.24	−0.55	<0.0001	0.006
l-Methionine	HMDB00696	0.43	0.06	0.47	<0.001	0.008
Unknown	—	0.44	−0.12	0.25	<0.001	0.009
d-Ribose	HMDB00283	0.67	0.19	0.47	0.003	0.009
d-Ribose	HMDB00283	−0.16	−0.68	−0.19	0.004	0.01
l-Methionine	HMDB00696	0	−0.67	0.05	0.005	0.01
l-Serine	HMDB00187	0.2	−0.11	0.38	<0.001	0.02
l-Serine	HMDB00187	0.23	−0.1	0.39	<0.001	0.02
l-Serine	HMDB00187	0.29	−0.03	0.41	0.001	0.02
l-Proline	HMDB00162	0.13	−0.41	0.07	0.001	0.02
l-Methionine	HMDB00696	0.34	−0.08	0.42	0.001	0.02
Glucosamine	HMDB01514	−0.15	−0.85	−0.53	0.01	0.02
Homocysteine	HMDB00742	0.26	−0.13	0.46	0.012	0.02
Unknown4	—	−0.08	−0.63	−0.11	0.014	0.02
Unknown5	—	−0.37	−0.75	−0.3	0.015	0.02
l-Valine	HMDB00883	0.18	−0.38	0.23	0.016	0.02
Unknown6	—	0.43	−0.43	−0.01	0.001	0.03
Citric acid	HMDB00094	0.28	−0.37	0.3	0.001	0.03
Unknown7	—	0.42	−0.45	−0.02	0.001	0.03
Unknown8	—	0.42	−0.45	−0.02	0.001	0.03
Urea	HMDB00294	0.49	0.63	0.74	0.001	0.03
l-Serine	HMDB00187	0.23	−0.09	0.39	0.001	0.03
Urea	HMDB00294	0.61	0.26	0.62	0.002	0.03
Unknown9	—	0.16	−0.09	0.14	0.002	0.03
l-Proline	HMDB00162	0.19	−0.32	0.14	0.002	0.04
Citric acid	HMDB00094	−0.48	−0.58	−0.77	0.034	0.04
Unknown10	—	0.34	−0.08	0.29	0.003	0.05
Citric acid	HMDB00094	−0.19	0.78	−0.19	0.003	0.05
Uric acid	HMDB00289	−0.02	0.07	−0.43	0.004	0.05
Urea	HMDB00294	0.48	0.1	0.7	0.007	0.09
l-Serine	HMDB00187	0.19	−0.05	0.36	0.008	0.1
Unknown11	—	−0.16	−0.16	0.39	0.008	0.11
Citric acid	HMDB00094	−0.02	−0.03	−0.65	0.009	0.11
l-Tyrosine	HMDB00158	−0.04	−0.03	−0.5	0.01	0.11
l-Tyrosine	HMDB00158	−0.22	−0.09	−0.4	0.01	0.11
Unknown12	—	0.6	0	0.47	0.01	0.11
l-Valine	HMDB00883	−0.44	−0.08	−0.34	0.012	0.13
l-Lysine	HMDB00182	−0.32	0.11	−0.18	0.013	0.13
Unknown13	—	0	0	0.49	0.013	0.13
Unknown14	—	0.46	0	0	0.014	0.13
Unknown15	—	0.45	0	0	0.014	0.13
4-Hydroxyproline	HMDB00725	0.77	0.74	−0.19	0.015	0.13
2-Hydroxybutyric acid	HMDB00008	0.42	0.06	0.45	<0.001	0.14
l-Methionine	HMDB00696	0.3	0	0.26	<0.001	0.14
Unknown16	—	0.31	−0.41	0	0.016	0.14

Altogether, 53 peaks (corresponding to 18 nonredundant metabolites) and 16 unknown peaks were identified as changing with a standard FDR of <0.15. Upon refining the data analysis, we discovered 34 peaks (corresponding to 12 nonredundant metabolites) and 9 unknown peaks with an FDR of 0.05 (Table 2). There was significant variation in the metabolite levels for the different treatments. At weaning, methionine was generally shown to be greatly increased in either of the ENS groups, reflecting the supplementation in the diet (Table 2). Significant changes for IUGR^mat/IUGR^pat lineage control-fed animals were decreased myoinositol, citric acid, l-proline, glucosamine, and l-valine, whereas urea was increased compared with Sham^mat/Sham^pat lineage control. The treatment of Sham^mat/Sham^pat lineage with ENS resulted in a decrease in glycine, whereas other amino acids and derivatives methionine, serine, urea, and hydroxybutyric acid all increased. Treatment of IUGR^mat/IUGR^pat lineage with an ENS diet increased myoinositol, methionine, citric acid, serine, l-proline, glucosamine, homocysteine, l-valine, and d-ribose, thereby largely reversing most of the decreasing metabolites observed among IUGR^mat/IUGR^pat lineage offspring fed a control diet (i.e., those destined for adult MetS and obesity). There were no significant differences between the sexes (Supplemental Table 1).

A similar trend was detected in adulthood (d160). In these experiments, the IUGR^mat/IUGR^pat lineage control-fed group had decreased expression of many metabolites compared with other groups. Similar to samples obtained from F2 offspring at weaning, the groups clustered independently. The consumption of ENS was, however, a greater determinant of dissimilarity when comparing IUGR^mat/IUGR^pat with Sham^mat/Sham^pat lineage grouping (Fig. 2B). Furthermore, in adulthood (d160), fewer metabolites were identified as a result of the use of a less-sensitive MS-GC for a single experiment (Table 3). Nevertheless, significant differences were detected that echoed our previous findings (40).

TABLE 3.

Metabolites and small molecule intermediates that significantly vary by virtue of heritable lineage and dietary exposure in utero and during lactation

Metabolite	Identification	Shammat/Shampat ENS	IUGRmat/IUGRpat Cont	IUGRmat/IUGRpat ENS	P	FDR
Malic acid	HMDB00744	0	2.14	0.07	0.0001	0.002
Myoinositol	HMDB00211	−0.25	1.74	0.87	0.02	0.1
l-Methionine	HMDB00696	0.43	1.99	0.28	0.03	0.1
Glycine	HMDB00123	−0.72	1.3	−0.44	0.03	0.1
l-Alanine	HMDB00161	0.88	2.12	1.05	0.04	0.11

To better visualize and measure significance of the metabolite changes, potentially significant, select metabolites were plotted as hive plots from replicate individual animals (Fig. 3). The Z score of the overall significantly changing peak mean is plotted on a separate axis for all peaks of a given metabolite. The metabolites plotted are indicated in bold in Table 2 and include the following: citric acid (a central component of oxidative phosphorylation; Fig. 3A), glucosamine (a monosaccharide precursor of proteoglycans and glycolipids; Fig. 3B), myoinositol (diverse functions in fat metabolism, gene expression, and signal transduction; Fig. 3C), and finally, the amino acid proline (Fig. 3D). For all 4 metabolites, there was a significant reduction in expression from Sham^mat/Sham^pat lineage control fed compared with IUGR^mat/IUGR^pat lineage control fed (ANOVA, Tukey’s test, P < 0.05; Fig. 3A–D). The only significant change from Sham^mat/Sham^pat lineage control fed to Sham^mat/Sham^pat lineage ENS fed was 1 of the 3 citric acid peaks. Interestingly, ENS appeared to confer a rescue of the levels in the IUGR^mat/IUGR^pat lineage ENS-fed cohort, with citric acid, glucosamine, myoinositol, and proline peaks exhibiting significantly increased expression compared with the IUGR^mat/IUGR^pat lineage control-fed group (ANOVA, Tukey’s test, P < 0.05). However, this rescue was not complete, as the comparison of IUGR^mat/IUGR^pat lineage ENS fed with Sham^mat/Sham^pat lineage control fed revealed significant differences in glucosamine and myoinositol expression (ANOVA, Tukey’s test, P < 0.05; Fig. 3B, C). These differences were largely preserved when comparing unprocessed raw spectra data. Glucosamine, myoinositol, and proline exhibit reduced levels significantly in IUGR^mat/IUGR^pat lineage control-fed compared with Sham^mat/Sham^pat lineage control-fed levels, whereas the effects of IUGR lineage and ENS treatment are apparent for citric acid (Table 4).

Figure 3.

Select markers of IUGR insult and later MetS that were altered significantly at weaning (d21). The metabolite expression Z scores are plotted for individual animals and connected to paired and matching m/z in neighboring groups in a hive plot. Variously shaded lines of purple indicate different m/z of the same metabolite. Mean peak intensity (Z = 0) is indicated by the gray line, whereas green and red lines are Z = 1 and −1, respectively. All peaks identified for citric acid (A), glucosamine (B), myoinositol (C), and proline (D) are plotted. Changing metabolites plotted are in boldface and thereby, noted as significant, as shown in Table 2.

TABLE 4.

Scaled raw peak intensity for significantly changing metabolites highlighted in Table 2

Metabolite	Peak	Shammat/Shampat Cont	Shammat/Shampat ENS	IUGRmat/IUGRpat Cont	IUGRmat/IUGRpat ENS
Citric acid	1	100 (7.9)	63.6 (22.7)	49.2 (1.3)a	39.1 (2.1)a
	2	100 (25.1)	282.6 (43.8)a	42.2 (27.8)	319.9 (52.3)b
	3	100 (39.5)	310.3 (47.1)b	17 (17.1)	339.9 (50.2)b
Glucosamine	1	100 (6.6)	86.9 (27.7)	18.9 (1.6)a	38.5 (10.2)
Myoinositol	1	100 (21.4)	71.0 (16.0)	10.8 (1.9)b	45.5 (7.2)
Proline	1	100 (7.7)	118.5 (9.6)	60.2 (6.4)a	111.5 (14.0)
	2	100 (6.3)	117.9 (7.5)	76.7 (5.8)	113.1 (8.3)

sem is indicated within in parentheses.

^aP < 0.05.

^bP < 0.01 compared to Sham^mat/Sham^pat control fed column (ANOVA with Dunnett’s).

The distinct effects of heritable MetS IUGR lineage and ENS diet supplementation were assessed further for significance by use of multivariate analysis. PLS-DA analysis revealed that IUGR^mat/IUGR^pat lineage control-fed animals separated and clustered distinctly from their Sham^mat/Sham^pat lineage control-fed counterparts long before adult MetS onset at d21 (P < 0.01; Fig. 4A). Of note, these animals had not yet been weaned, so their serum markers are indicative of gestational and lactational exposures. Furthermore, IUGR^mat/IUGR^pat lineage ENS fed separated along the horizontal axis in the opposite direction, which indicates that ENS and IUGR have opposing effects at weaning. Interestingly, Sham^mat/Sham^pat lineage groups that did not develop MetS and IUGR^mat/IUGR^pat lineage rats rescued from the MetS phenotype through ENS diet clustered together in adulthood (Fig. 4B); rats with MetS (Fig. 1C; IUGR^mat/IUGR^pat-Cont) separated distinctly from the other 3 cohorts, similar to d21 (P < 0.05; Fig. 4B). The separation of the IUGR^mat/IUGR^pat lineage control-fed group is demonstrated further by randomForest analysis (class error < 0.1; Fig. 5). Altogether, this indicates that there is a separation in the overall metabolic profile of the d21 weanlings destined for metabolic disease before the actual onset of MetS in adulthood.

Figure 4.

PLS-DA regression analysis of serum metabolite levels. Data representative of 3 independent experiments at the time of weaning (d21) are shown (A), alongside that from adult animals at d160 (B). IUGR-Cont diet clustered separately at both time-points, whereas clustering was by the presence or absence of MetS phenotype by d160 (B). Permutation analysis revealed the clustering to be significant (P < 0.01 and P < 0.05 for A and B, respectively). A 95% confidence area is indicated for groups.

Figure 5.

randomForest analysis for group changes. At the time of weaning on d21, IUGR lineage control fed could be differentiated from other treatment groups a minimum of 9 times out of 10 (out of bag error, y axis).

DISCUSSION

We have shown previously that MetS is heritably transmitted along the maternal and paternal lineage from the P1 (grandmaternal) to F2 generation following bilateral uterine artery ligation of only the P1. Moreover, we have shown that adult MetS was greatly ameliorated following supplementation of essential nutrients along the 1-carbon pathway (40). In this current study, we significantly and meaningfully expanded our initial findings by use of MS on an unbiased platform to identify underlying changes in metabolites and small molecular intermediates. First, we show that the serum metabolome in the F2 IUGR^mat/IUGR^pat lineage is perturbed at the time of weaning, long prior to the onset of MetS. Second, we found that dietary supplementation of the F1 generation with 1-carbon intermediates partially normalized the F2 IUGR-mediated metabolite changes at weaning (d21), and by adulthood (d160) there was a near complete reversal of the F2 metabolite perturbations. Of particular interest, the altered metabolite profiles of citric acid, glucosamine, myoinositol, and proline in the F2 IUGR^mat/IUGR^pat lineage were predictive of the onset of MetS. Collectively, these findings suggest that the underlying metabolome associated with the development of MetS is inherited and disturbed long before the manifestation of the adult phenotype. Moreover, the reversal of the inherited metabolomic profile in the F2 IUGR^mat/IUGR^pat lineage rats with ENS diet indicates that heritable metabolic disease fate is amenable to dietary intervention.

In a broad sense, the metabolites identified as changing were primarily amino acids or small molecule intermediates related to the citric acid cycle and carbohydrate metabolism. As with humans, the only source of methionine in rats is dietary. As assurance that ENS was consumed, methionine, a direct methyl donor in DNA methylation and a major ENS component, was repeatedly and consistently identified as elevated in ENS-treated groups but not control fed. Therefore, its elevated serum levels with ENS were taken as an indication of the validity of the metabolomic analysis. In agreement, serine is converted to glycine through a reaction with methionine tetrahydrofolate, a methionine derivative (48). Serine is also consumed in the conversion of methionine to cysteine via homocysteine (48). As such, the changes observed in these 2 aa may have some overlapping influence from the supplementation. However, independently metabolized amino acids were also altered. For instance, valine was changed significantly with IUGR^mat/IUGR^pat lineage and ENS feeding. As an essential amino acid, altering levels of valine could only result from differing rates of use, which indicates potential changes in metabolism (49–54).

Four metabolites of particular interest were citric acid, glucosamine, myoinositol, and proline, as these metabolites were remarkable in that their expression appeared to be in close agreement with the later development of MetS. Along these lines, all 4 metabolites were decreased significantly for IUGR^mat/IUGR^pat lineage control fed, whereas IUGR^mat/IUGR^pat lineage ENS fed had metabolite levels similar to Sham^mat/Sham^pat lineage controls. Therefore, the expression pattern of these metabolites reflects the later development of MetS long before the development of the adult phenotype. Thus, these metabolites may act as predictive, early markers of MetS.

Some of the metabolites are directly linked to carbohydrate homeostasis and potentially contribute to the manifestation of adult metabolic diseases, as they are implicated in energy metabolism. Citric acid is a central molecule in ATP production and has been linked to obesity and insulin resistance. Mechanistic analyses have linked insulin resistance to dysfunction of oxidative phosphorylation in the mitochondria (55, 56). Many of the genes coding for enzymes downstream of citric acid, including aconitase 2, isocitrate dehydrogenase 2, and succinate-coenzyme A (CoA) ligase, α subunits 1 and 2, whose translated products collectively convert citric acid to succinyl-CoA, appear up-regulated in the subcutaneous fat of patients predisposed to insulin resistance. Conversely, gene products that convert succinyl-CoA to succinate, including succinate dehydrogenase complex, subunits A–D, are relatively unchanged or down regulated (57). Therefore, it is possible that succinyl-CoA or derived downstream metabolites are accumulating at the expense of citric acid in the setting of insulin resistance [which we have demonstrated to occur in our adult animals (40)]. This could lead to excessive metabolites that may contribute to hypercholesterolemia, which was observed in our rats (40, 58, 59). Furthermore, insulin resistance is known to result in increased hepatic gluconeogenesis, which may also deplete citric acid and other citric acid cycle intermediaries.

Myoinositol is a cyclitol essential for proper cellular function and has been implicated in glucose homeostasis. Altered fetal levels of myoinositol have been shown in association with IUGR in humans and animal models (60–63). Aberrations of myoinositol metabolism are associated with insulin resistance and diabetic complications [reviewed in ref. (64)], where its action is as a second messenger downstream of insulin signaling. Myoinositol is synthesized downstream from the glucose-6-phosphate precursor, whose levels are elevated in the diabetic rat. Exogenous myoinositol improves insulin sensitivity in mice (65) and lowers blood glucose levels in insulin-resistant rhesus monkeys (66). Moreover, supplementation with myoinositol was able to ameliorate the MetS phenotype with improvements in serum glucose, insulin, and lipid levels and lowered blood pressure (67). It also halved the risk of gestational diabetes in an at-risk cohort (68). Interestingly, metabolomic analyses of neonatal pigs have identified a correlation between birth weight and myoinositol levels, which suggests one means by which a predisposition to MetS may be influenced by the in utero environment (69). Thus, our present findings of decreased levels in IUGR^mat/IUGR^pat lineage control-fed rats may be reflective of altered sensitivity to insulin and early predisposition to MetS, which is consistent with previous findings (60, 67, 70).

The metabolomic changes that occur as a result of the grandmaternal uterine artery ligation (uteroplacental insufficiency) ultimately arise concomitant with the conditions of hypoinsulinemia, hypoglycemia, acidosis, and hypoxia that manifest in utero (71, 72). Alterations associated with intrauterine hypoxia following uterine artery ligation include key glucose-regulating processes, such as glucose transporter 1 and 2 activity (73), altered mitochondrial function (74, 75), and insulin-mediated regulation (13, 40), as well as nephropathy (76, 77) and hepatic regulatory changes (78–80), which are critical to metabolic function. Epigenetic regulation of key intermediates to metabolic development and function, including IGF-1, peroxisome proliferator-activated receptor-γ coactivator 1a, carnitine palmitoyltransferase 1, and pancreatic and duodenal homeobox 1, have been demonstrated to occur in concert with uteroplacental insufficiency-induced IUGR (30, 32, 49, 81). It is possible that heritable modifications in gene methylation may explain the heritable metabolome changes observed herein, as we have published previously (40). Along these lines, human studies obtained from historical data, such as the Dutch famine, concur that molecular and physiologic changes related to metabolism exert effects across generations (4, 14, 22, 82)

Maternal dietary availability of cofactors in 1-carbon metabolism and the conversion of S-adenosylmethionine to S-adenosylhomocysteine have been shown to influence CpG methylation and disease (37, 83, 84). Additionally, studies of seasonal abundances in agrarian cultures in Gambia have revealed the ability of relative dietary abundance of essential 1-carbon metabolism nutrients to alter metastable allelic methylation in human populations (38, 39). Our ENS diet contained 1-carbon pathway substrates, cofactors, and intermediaries methionine, choline, betaine, B₁₂, zinc, and folate (85). It is possible that ENS supplementation of key components of 1-carbon metabolism [many of which are reduced following IUGR (29)] influences site-specific DNA methylation of key genes in crucial regulatory regions, resulting in altered expression and the subsequent abrogation of adult disease in our model. This view is buttressed by our own published findings of the model here employed revealed that the proximal promoters of IGF-1, a key regulator of metabolism and growth, is epigenomically altered both in gestation and postnatally in response to IUGR and reverts with exposure to an ENS diet (40). We speculate that an additional mechanism of ENS influence may be through the lasting influence of an improved perfusion of the placenta in the F1 pregnancies via eNOS-mediated vasodilation, which makes use of arginine as a substrate in NO production. Arginine supplementation in pregnancy has shown some beneficial effects for offspring in animal models (86, 87); however, its effect on placental transport and function has yet to be characterized.

Our current findings are buttressed by the large number of samples analyzed, by use of robust statistical and analytic techniques, and by use of a model system replicative of current, common causes of IUGR in humans. The use of uterine artery ligation models is considered the best model of in utero restriction, as it most closely resembles the gestational starvation of uteroplacental insufficiency [a result of maternal hypertensive comorbidities, abnormal placentation, environmental exposures, such as maternal tobacco use, or idiopathic angiogenic failure (88–90)], which, rather than malnutrition, is a common etiological factor of IUGR. Ergo, the uterine artery ligation model, may offer some translational advantages over caloric or protein-restriction models, and therefore, this model was chosen in an effort to recapitulate the conditions of acidosis and hypoxia that accompany uteroplacental insufficiency in gestation (71, 72). However, the findings here are subject to the same caution as many other studies of multigenerational models. Germline inheritance is a demonstrated phenomenon (91), and it remains a possibility that the F1 gametes exposure to IUGR during meiosis of the developing oocytes is a factor in F2 changes. Likewise, bioactive compounds in the germ cells may be contributing to the altered phenotype. It is also possible that subtle metabolic alterations in F1, which had not yet developed into a diseased phenotype, may have contributed to the F2 phenotype preweaning. Although the interpretation of the significance is limited by the methodology’s incompatibility with determining absolute changes in levels, nevertheless, broad quantitative metabolite profiling demonstrated that for at least 2 generations, there is a transfer of an adverse phenotype that manifests clearly in early life and that also persists into adulthood after the removal of the F1 influence.

There are numerous metabolomic analyses of the immediate effects of IUGR, and on later metabolism, however, fewer studies have examined metabolomic changes in concert with interventions that modify the onset of adult MetS. Specific studies have examined the effects of IUGR in cord blood or urine immediately following birth (60, 62, 92–95). Others, by use of animal models, have examined the effects of IUGR (96–99), as well as developing MetS (60, 97, 100), in the F1 generation. To our knowledge, we are the first to demonstrate a broadly changing metabolome that is multigenerationally inherited. Furthermore, we were able to identify putative serum biomarkers associated with adult onset of inherited MetS, and we were able to rescue the MetS-associated metabolome by dietary supplementation (40). Therefore, implications are 3-fold. First, inherited MetS is preceded by metabolic perturbations before the influence of diet at weaning. Second, small molecule metabolites may predict risk of MetS long before its manifestation. Finally, nutritional interventions, which mitigate these adult morbidities, may abrogate some of these metabolite alterations. In aggregate, our findings, to date, provide initial evidence that heritable fetal growth restriction, resulting in adult MetS, is preceded by serum alterations in metabolites that are, at least partially, reversible by nutritional supplementation. These nutritional interventions may yield clues into crucial and modifiable pathways, and we speculate that future investigations targeting these pathways may provide innovative means for modifying the prevalence of developmentally programmed adult metabolic disease.

Glossary

CoA

coenzyme A

d

postnatal day

e

embryonic day

ENS

essential nutrient supplementation

FDR

false discovery rate

GC-MS

gas chromatography-mass spectrometry

IUGR

intrauterine growth restriction

IUGR^mat

maternal IUGR

IUGR^pat

paternal IUGR

MetS

metabolic syndrome

m/z

mass-to-charge

PLS-DA

partial least-squares discriminate analysis