Maternal gestational obesity is associated with elevated risks for neurodevelopmental disorder, including autism spectrum disorder. However, the mechanisms by which maternal adiposity influences fetal developmental programming remain to be elucidated. We aimed to understand the impact of maternal obesity on the metabolism of both pregnant mothers and their offspring, as well as on metabolic, brain, and behavioral development of offspring by utilizing metabolomics, protein, and behavioral assays in a non-human primate model. We found that maternal obesity was associated with elevated inflammation and significant alterations in metabolites of energy metabolism and one-carbon metabolism in maternal plasma and urine, as well as in placenta. Infants born to obese mothers were significantly larger at birth compared to those born to lean mothers. Additionally, they exhibited significantly reduced novelty preference and significant alterations in their emotional response to stress situations. These changes coincided with differences in phosphorylation of enzymes in the brain mTOR signaling pathway between infants born to obese and lean mothers and correlated with the concentration of maternal plasma betaine during pregnancy. In summary, gestational obesity significantly impacted the infant systemic and brain metabolome and adaptive behaviors.

2.1 Study population

Animal handling was approved by the UC Davis Institutional Animal Care and Use Committee (IACUC protocol#19299) and all experiments were performed in accordance with relevant guidelines and regulations.

Pregnant female rhesus macaques (Macaca mulatta) with appropriate social behavior and a previous successful pregnancy were selected from an indoor breeding colony at the California National Primate Research Center (CNPRC). Fetal sex was determined by a qualitative real time PCR assay to detect the Y chromosome and only those with male fetuses were chosen for this study. Animals used in this study had maintained a consistent BCS for at least one year prior to the study. Obesity is defined when subjects have body fat above 30% for women (American Medical Association). A BCS of 3.5 was chosen as the cut off for inclusion in the obese group for this study as a BCS of 3.5 is correlated with 32 % body fat [14]. Mothers with a BCS of 2.5 or lower comprised the Lean group. The sample size of the biological samples was not balanced due to fetal deaths for unknown reasons, misidentification of female offspring, technical issues in collecting enough sample volume for analysis, or recruitment of additional animals into the study in the middle of pregnancy to account for the sample loss (Supplementary Table S1). The final number of mothers and their offspring included six for the Lean and seven for the Obese groups (Supplementary Table S2).

2.2 Feeding and rearing of animals

Adult animals were fed seven biscuits (#5047; LabDiet, St. Louis, MO, USA) twice daily between 6-9 am and 1-3 pm. All mothers were provided nine biscuits twice daily once pregnancy was determined, and twelve biscuits twice daily while nursing infants 4 months and older. Fresh produce was provided biweekly and water was available ad libitum. More detailed description is available in Appendix A.

2.3 Sample collection and processing

All animals were coded by IDs, and therefore, the animal care takers and researchers who collected samples were blinded for group assignment. The collected biological samples were randomized using random numbers generated in R in conducting assessments, and the group assignment was blinded until the data was analyzed. On the day prior to sample collection, food was removed approximately 30 min after the feeding in the afternoon, and biological samples were collected before the morning feeding. Pregnancy in rhesus macaques lasts for 166.5 days on average. Plasma and urine samples were collected from mothers once during the 1^st and 2^nd trimesters, and twice during the 3^rd trimester on GD45, 90, 120, and 150 after anesthetizing animals with 5-30 mg/kg ketamine or 5-8 mg/kg telazol. Blood samples were collected in 5 mL lavender top (EDTA) or green (heparin) tubes and the supernatant was collected. Urine was collected from the bladder by ultrasound-guided transabdominal cystotomy using a 22-gauge needle and subsequently centrifuged to collect supernatant. Within 15 days prior to delivery, a placental sample was collected transabdominally under ultrasound guidance using an 18-gauge needle attached to a sterile syringe and centrifuged to collect the pellet.

Infant plasma was collected at PD30, 90, and 110, and plasma, urine, and brain tissues were collected at PD180 when necropsy was conducted between 9:30 am-1:30 pm. Infants were anesthetized with ketamine and plasma and urine were collected, followed by euthanasia with 120 mg/kg pentobarbital. Heparin injection and clamping of the descending aorta was followed by flushing with saline at room temperature for 2 min and then by saline at 4 °C for 5 min at 250 mL/min until clear. The brain was extracted, and four regions (amygdala, hippocampus, hypothalamus, and prefrontal cortex) were dissected and immediately frozen. All collected samples were stored at -80 °C.

2.4 Metabolite extraction and insulin, cytokine, and cortisol measurement

The plasma and urine samples were thawed on ice and filtered by Amicon Ultra Centrifugal Filter (3k molecular weight cutoff; Millipore, Billerica, MA, USA), and the filtrate was used for metabolomics analysis. Samples were stored at 4 °C overnight and pH was adjusted to 6.8 ± 0.1. For the placental samples, polar metabolites were extracted as described with the following modification: after lyophilization of the polar metabolite layer, the dried sample was reconstituted in 270 µL of 10 mM phosphate buffer (pH 6.85) prepared in deuterium oxide. Samples were transferred to 3 mm Bruker NMR tubes and kept at 4 °C until NMR data acquisition within 24 hours of sample preparation.

A multiplex Bead-Based Kit (Millipore) was used to measure insulin as well as 17 cytokine and chemokine levels in heparin-treated plasma samples including hs-CRP, Granulocyte-macrophage colony-stimulating factor (GMCSF), interferon-γ (IFN-γ), tumor necrosis factor-α (TNF-α), transforming growth factor-α (TGF-α), monocyte chemoattractant protein-1 (MCP-1), macrophage inflammatory protein-1β (MIP-1β), interleukin (IL)-1β, IL-1 receptor antagonist (IL-1ra), IL-2, IL-6, IL-8, IL-10, IL-12/23 p40, IL-13, IL-15, and IL-17A. The assay was performed following the manufacturer’s protocol. Assessment of infant plasma cortisol was conducted as previously described [17,18]. Briefly, infants were separated from their mothers at 9 am and blood samples were collected at 11 am (Sample 1). Another blood collection was done at 4 pm (Sample 2), followed by intramuscular injection of 500 μg/kg Dex. Blood was collected at 8:30 am of the following day (Sample 3) and 2.5 IU of ACTH was then injected intramuscularly. After 30 min, the last blood was collected (Sample 4). More detailed description is available in Appendix A.

2.5 ¹H nuclear magnetic resonance (NMR) spectroscopy data acquisition

We conducted an untargeted metabolomics analysis using ¹H NMR spectroscopy. All spectra were acquired at 25 °C using the noesypr1d pulse sequence on a Bruker Avance 600 MHz NMR spectrometer (Bruker, Billerica, MA, USA). Identification and quantification of metabolites were completed using Chenomx NMRSuite (version 8.1, Chenomx Inc., Edmonton, Canada).

2.6 Protein analysis

Protein was extracted from the cell layer collected after metabolite extraction of brain samples, and protein quantification and western blots were done as previously described. The following antibodies from Cell Signaling Technology (Danvers, MA, USA) were used for the western blots: rabbit anti-Akt (#9272), anti-phospho-Akt (#9275; Thr308), anti-AMPKα (#2603), anti-phospho-AMPKα (#2535; Thr172), anti-p70 S6K (#9202), anti-phospho-p70 S6K (#9234; Thr389), as well as goat anti-rabbit IgG antibody conjugated to horseradish peroxidase (#7074). Either Clarity Western ECL Blotting Substrates (Bio-Rad) or Radiance Plus (Azure biosystems, Dublin, CA, USA) were used depending on the strength of signal for chemiluminescent detection. Refer to Appendix A for more detailed description.

2.7 VPC Test

Recognition memory was tested with infants on post-conception day 200 ± 3 days using a VPC test conducted between 8:30-10:30 am. Briefly, infants were hand-held in front of a testing booth to look at two identical black and white high contrast abstract pictures (Fagan Test of Infant Intelligence; Infantest Corporation, Cleveland, OH, USA) placed to the right and left of center for a total of 20 sec of cumulative looking time (familiarization trial). Then, the familiar and novel pictures were placed either right or left of center according to a pre-decided random order for a 10 sec test period from the time of the first fixation (preference trial 1). The side of the pictures was switched and a second 10 sec test period was conducted (preference trial 2). Four problems were presented to each infant. The trials were video recorded for later scoring of frequency and duration of looking patterns using The Observer software (Noldus, Inc., Wageningen, The Netherlands). Novelty preference was calculated as: number of fixations at the novel stimulus/number of fixations at both the novel and familiar stimulus.

2.8 HI test

The HI test was conducted as described previously. In short, we examined the frequency of scratch (as an indicator of anxiety) in response to the following four graded levels of stress (1 min each): Profile-Far (technician presented the left profile from ~1 m away from an infant in a cage), Profile-Near (presented left profile from ~0.3 m), Stare-Far (made direct eye contact with the animal from far), and Stare-Near (direct eye contact from near position).

2.9 Statistics

The overall gestational weight gain (GWG) rate was obtained using the following equation: 1000 * ln(W2/W1)/(D2-D1), where D1 is the date the last weight (W1) was obtained before pregnancy and D2 is the date the last weight (W2) was obtained before birth. Homeostatic model assessment for insulin resistance (HOMA-IR) was calculated using fasting glucose (mg/dL) * fasting insulin (μU/mL)]/405. Estimated placental volume (EPV) was calculated as follows: (πT/6) * [4H(W−T)+W(W−4T)+4T², where T is thickness at maximal height, H is height at maximal width, and W is maximal width measured by ultrasound.

Metabolite concentrations were log-transformed prior to application of the following statistics. A linear mixed-effects model was fitted followed by ANOVA to test the group difference of samples with multiple time points using lme4 package (version 1.1.21). Estimated marginal means were obtained using emmeans (version 1.4.4), and pairwise comparison was applied as a post-hoc test. R² values were obtained as the effect size using r2glmm (version 0.1.2) (medium effect when 0.15<R²<0.35; large when 0.35 R²). T-tests were used to test for group differences of samples collected at a single time point. The effect size was calculated using Cohen’s D (d) measurement (medium effect when 0.5<d<0.8; large when 0.8 d<1.3; very large when 1.3 d). For metabolomics analysis, Benjamini-Hochberg false discovery rate correction was applied to p-values. For post-hoc tests, a p-value of 0.05 was used as the cut-off. Correlation between HOMA-IR and EPV was assessed by repeated measures correlation using rmcorr package (version 0.4.1) and other correlations were tested by the Pearson or Spearman methods and visualized using ggscatter (version 0.2.5). The correlation coefficient (R) was used as the effect size (medium effect when 0.3<R<0.5; large when 0.5 R 1). In order to address the small sample size and the inherent large heterogeneity and variability of rhesus macaque data, we utilized a combination of p-values < 0.1 and medium-large effect size to define statistical significance. More detailed description is available in Appendix A.