Maternal obesity has been associated with a higher risk of pregnancy-related complications in mothers and offspring; however, effective interventions have not yet been developed. We tested two common interventions, calorie restriction and pravastatin administration, during pregnancy in a rhesus macaque model with the hypothesis that these interventions would normalize metabolic dysregulation in pregnant mothers leading to an improvement in infant metabolic and cognitive/social development. A total of 19 obese mothers were assigned to either one of the two intervention groups (n=5 for calorie restriction; n=7 for pravastatin) or an obese control group (n=7) with no intervention, and maternal gestational samples and postnatal infant samples were compared with lean control mothers (n=6). Gestational calorie restriction normalized one-carbon metabolism dysregulation in obese mothers but altered energy metabolism in their offspring. Although administration of pravastatin during pregnancy tended to normalize blood cholesterol in the mothers, it potentially impacted the gut microbiome and kidney function of their offspring. In the offspring, both calorie restriction and pravastatin administration during pregnancy tended to normalize the activity of AMPK in the brain at 6 months, and while results of the Visual Paired-Comparison test, which measures infant recognition memory, were not significantly impacted by either of the interventions, gestational pravastatin administration, but not calorie restriction, tended to normalize anxiety assessed by the Human Intruder test. Although the two interventions tested in a non-human primate model led to some improvements in metabolism and/or infant brain development, negative impacts were also found in both mothers and infants. Our study emphasizes the importance of assessing gestational interventions for maternal obesity on both maternal and offspring long-term outcomes.

Study population

Pregnant female rhesus macaques (Macaca mulatta) from an indoor breeding colony at the California National Primate Research Center with appropriate social behavior and previous successful pregnancies were enrolled. Animal handling was approved by the UC Davis Institutional Animal Care and Use Committee (IACUC) (#19299). A qualitative real-time PCR assay (Jimenez & Tarantal, 2003) was used to identify mothers with male fetuses to include in this study. Since obesity is defined as subjects with body fat above 30% for women, according to guidelines from the American Society of Bariatric Physicians, American Medical Association, and in some publications (Okorodudu et al., 2010; Shah & Braverman, 2012), a Body Condition Score (BCS) of 3.5 (32.8 % body fat on average (Summers et al., 2012)) was used as the cutoff. Therefore, mothers with BCS of 3.5 and above were categorized as obese. Obese mothers were randomly assigned to the Obese Control (OC) group, OR group (received calorie Restriction), or OP group (received Pravastatin). Mothers with BCS of 2.5 and below were assigned to the Lean Control (LC) group. The unbalanced sample size was because some mothers were removed from the analyses due to fetal deaths for unknown reasons, misidentification of a female fetus, different timing for study enrollment, or technical issues upon collecting samples. The number of animals was six for the LC, seven for the OC, five for the OR, and seven for the OP groups.

Feeding, rearing, and interventions

Adult female animals were provided monkey diet (High Protein Primate Diet Jumbo #5047; LabDiet, St. Louis, MO, USA) twice a day between 6–9 am and 1–3 pm. The calories were provided as 56% from carbohydrates, 30% from protein, and 13% from. Mothers in the LC, OC, and OP groups were fed nine biscuits twice a day once pregnancy was confirmed. Mothers in the OR group received a restricted supply of food once the pregnancy was detected and was maintained throughout pregnancy. The food restriction was set such that the average total weight increase would be 8% body weight from the last day before conception because the recommended total weight gain in the 2^nd and 3^rd trimesters is 5-9 kg for the average US woman with obesity who weighs 80 kg and is 1.6 m in height (Body Mass Index of 30), according to the Institute of Medicine 2009 guidelines (Institute of Medicine and National Research Council, 2009). During nursing of infants older than 4 months, all mothers were provided twelve biscuits. Fresh produce was provided biweekly, and water was provided ad libitum for all mothers. Mothers in the OP group were given pravastatin sodium (ApexBio Technology, Houston, TX, USA) at 20 mg/kg body weight prepared in a neutralized syrup (20 mg/mL sodium bicarbonate dissolved in a fruit-flavored syrup (Torani, San Leandro, CA, USA)) once a day from the time pregnancy was confirmed until delivery. The caloric value of the administration was made so as not to influence body weight or skew nutritional value of the diet among all treatment groups. Both interventions were applied only during gestation. Although most mothers were allowed to deliver naturally, cesarean delivery was performed for fetal indications when recommended by veterinarians (2 for each of the LC and OC groups, and 1 for the OP group). These mothers did not accept their infant following birth, so foster mothers were provided.

Sample Collection and pre-processing prior to sample storage

The animal caretakers and researchers who collected samples were blinded for group assignment by coding all animals by IDs. The collected biological samples were randomized by using random numbers and the group assignment was blinded during the data collection.

Both mothers (during pregnancy) and infants were weighed every week. One day before sample collection, food was removed 30 min after the afternoon feeding, and biological samples were collected prior to the morning feeding. To collect biological samples, animals were anesthetized using 5–30 mg/kg ketamine or 5–8 mg/kg telazol. Both maternal and infant blood was collected using 5 mL lavender top (EDTA) tubes (Monoject, Cardinal Health, Dublin, OH, USA) and urine was collected from the bladder by ultrasound-guided transabdominal cystotomy using a 22-gauge needle and stored in a 15 mL Falcon tube. A placental sample was collected at GD150 transabdominally under ultrasound guidance using an 18-gauge needle attached to a sterile syringe. Sample processing was as previously described in (Hasegawa et al., 2022). Necropsy was conducted between 9:30 am–1:30 pm. First, infants at the age of PD180 were fasted and anesthetized with ketamine, and plasma and urine were collected. Then, euthanasia was performed with 120 mg/kg pentobarbital, followed by heparin injection, clamping of the descending aorta, and flushing with saline until clear. The kidney and brain (amygdala, hippocampus, hypothalamus, and prefrontal cortex) were collected, weighed, and immediately frozen on dry ice or liquid nitrogen to store at -80 °C until further analyses.

Metabolite extraction and analysis by ¹H NMR, and measurement of insulin, cholesterol, cytokine, and cortisol

Detailed procedures were previously described (Hasegawa et al., 2022). Briefly, plasma and urine samples were filtered using Amicon Ultra Centrifugal Filter (3k molecular weight cutoff; Millipore, Billerica, MA, USA), and the supernatant was used for analysis. For both the placental and brain tissue samples, polar metabolites were extracted using our previously reported method (Hasegawa et al., 2020). A total of 180 μL of sample (tissue extract or filtered urine or serum) was transferred to 3 mm Bruker NMR tubes (Bruker, Billerica, MA, USA). Within 24 h of sample preparation, all ¹H NMR spectra were acquired using the noesypr1d pulse sequence on a Bruker Avance 600 MHz NMR spectrometer (Bruker, Billerica, MA, USA) (O’Sullivan et al., 2013). Chenomx NMRSuite (version 8.1, Chenomx Inc., Edmonton, Canada) (Weljie et al., 2006) was used to identify and quantify metabolites.

Heparin-treated plasma samples were used to measure insulin and 17 cytokines and chemokines (hs-CRP, Granulocyte-macrophage colony-stimulating factor, IFN-γ, TNF-α, transforming growth factor-α, monocyte chemoattractant protein-1, macrophage inflammatory protein-1β (MIP-1β), and 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) using a multiplex Bead-Based Kit (Millipore) on a Bio-Plex 100 (Bio-rad, Hercules, CA) following the manufacturer’s protocol. For each sample, a minimum of fifty beads per region were collected and analyzed with Bio-Plex Manager software using a 5-point standard curve with immune marker quantities extrapolated based on the standard curve. Two samples were removed for analysis of TNF-α and IL-1ra as technical errors (both from Animal ID 1132103: 895.2 and 1115.1 pg/mL at gestational days (GD) 90; 510.8 and 617.2 pg/mL at GD120, respectively). Plasma cholesterol level was measured by Clinical Laboratory Diagnostic Product (OSR6116) on Beckman Coulter AU480 (Beckman Coulter, Brea, CA).

Infant plasma cortisol level at PD110 was assessed as previously described (Vandeleest et al., 2019; Walker et al., 2018). In short, infants were transferred to a test room at 9 am and blood was drawn at 11 am (Sample 1), followed by another blood collection at 4 pm (Sample 2) and intramuscular injection of 500 μg/kg dexamethasone (Dex) (American Regent Laboratories, Inc., Shirley, NY). On the next day, a blood sample was collected at 8:30 am (Sample 3), and then 2.5 IU of adrenocorticotropic hormone (Amphastar Pharmaceuticals, Inc., Rancho Cucamonga, CA) was injected intramuscularly. The last blood was collected (Sample 4) 30 min after adrenocorticotropic hormone injection. The collected blood samples were processed and stored, and cortisol concentration was assessed by a chemiluminescent assay on the ADVIA Centaur CP platform (Siemens Healthcare Diagnostics, Tarrytown, NY, USA).

Protein quantification and western blot analysis

Protein quantification and western blot analysis were conducted as previously described (Hasegawa et al., 2020). Briefly, a DC Protein Assay kit (Bio-Rad) was used for total protein quantification. For the western blot, the following antibodies were used: rabbit anti-Akt (#9272), anti-phospho-Akt (#9275; Thr308), anti-Adenosine Monophosphate-activated Protein Kinase (AMPK)α (#2603), anti-phospho-AMPKα (#2535; Thr172), anti-p70 S6 Protein Kinase (S6K) (#9202), anti-phospho-p70 S6K (#9234; Thr389), and goat anti-rabbit IgG antibody conjugated to horseradish peroxidase (#7074) (Cell Signaling Technology, Danvers, MA, USA). For chemiluminescent detection, Clarity Western ECL Blotting Substrates (Bio-Rad) or Radiance Plus (Azure biosystems, Dublin, CA USA) were used.

Visual Paired Comparison Test at ~1 month of age

On post-conception day 200 ± 3 days, recognition memory development of infants was assessed using Visual Paired Comparison (VPC) conducted between 8:30–10:30 am (Burbacher & Grant, 2012; Golub et al., 2006, 2014). In brief, two identical black and white contrast abstract pictures (Fagan Test of Infant Intelligence; Infantest Corporation, Cleveland, OH) were placed right and left of center, and infants were allowed to observe for a total of 20 sec (familiarization trial). Then, the pictures were switched to the familiar and novel ones placed either right or left of center based on a predetermined random order, and the frequency and duration of looking were video recorded using The Observer software (Noldus, Inc., Wageningen, The Netherlands) for a 10 sec from the time of the first fixation (preference trial 1). The side of the pictures was switched, and another 10 sec test period was recorded (preference trial 2). In total, four problems were conducted to each infant. Novelty preference was calculated by dividing the number of fixations at the novel stimulus by number of fixations at both the novel and familiar stimulus.

Human Intruder (HI) test at ~4 months of age

HI test was conducted as described before (Walker et al., 2018). Briefly, we measured the frequency of scratch (as an indicator of anxiety (Troisi, 2002)) for 1 min in four graded levels of stress: Profile-Far (the left profile of a technician was presented from ~1 m away from an infant), Profile-Near (left profile presented from ~0.3 m), Stare-Far (direct eye contact was made with the infant from far position), and Stare-Near (direct eye contact made from near position).

Statistical analyses

All statistical analyses were done in R (version 3.6.1). GWG was obtained using the following equation: [(delivery weight) - (last weight before conception)] / (last weight before conception) * 100, where the delivery weight is the last weekly weight before delivery. Homeostasis model assessment for insulin resistance (HOMA-IR) was obtained using the following equation: [fasting glucose (mg/dL) * fasting insulin (μU/mL)] / 405. kidney tissue weight was divided by the body weight at PD7 to obtain the normalized kidney mass. One animal with ID1215218 (OC group) was removed due to a technical error in measurement of insulin, as the insulin level was higher than biologically reasonable (Zhang et al., 2011). The following equation was used to calculate estimated placental volume (EPV): (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 (Sonek et al., 2018).

Either the OR or the OP group was separately compared with both control groups (OC and LC). For data that were collected at a single time point, one-way ANOVA was performed and Tukey honestly significant difference as a post-hoc test was applied. Eta squared was calculated as the effect size measurement using sjstats package (version 0.18.0), with a medium effect defined as 0.06< eta squared <0.14, and a large effect defined as eta squared ³ 0.14. For data collected at multiple time points, a linear mixed-effects model was fit followed by ANOVA to test for group differences using the lme4 package (version 1.1.21). Estimated marginal means were calculated by using emmeans (version 1.4.4) and pairwise comparison was applied as a post-hoc test. Effect size for multiple time points was calculated as coefficient of determination (R²) using r2glmm (version 0.1.2), with a medium effect defined as 0.15<R²<0.35, and a large effect defined R² >=0.35.

For metabolomics data, metabolite concentrations were log-transformed prior to statistical analysis. Benjamini-Hochberg false discovery rate for multiple testing correction was applied. After testing for the model fit, the statistical model was adjusted by maternal age for maternal plasma samples, by the centered GWG values for maternal urine samples, and by maternal age for infant plasma samples. Infant weights were corrected for gestational length. In order to avoid false negative results due to the small sample size, statistical significance was defined as a combination of uncorrected p-value of less than 0.1 and medium-large effect (Sullivan & Feinn, 2012). Unless otherwise stated, data are presented as mean ± standard error. We used duration of looking from eye tracking test measured at months 0.25, 1, 3, and 6 as the primary outcome to calculate the power. Based on the previous study (Bauman et al., 2013), we assumed a correlation among the four repeated observations ranges between 0.3 and 0.7, for a type one error α = 0.05, the intended sample size of eight monkeys per group can provide a power of 80-93% to detect the rates of change per month between the two groups as small as 1.1 SD (Diggle et al., 2002).

For cytokine/chemokine data, Partial Least Squares Discriminant Analysis (PLS-DA) was performed as well to examine whether different combinations of multiple cytokines could be used to differentiate between treatment groups. The PLS-DA model was used to cluster the groups and the Variable Importance in Projection (VIP) scores were obtained measure the importance of cytokine/chemokine in the PLS-DA model. Cytokine/chemokines with VIP scores above 1.0 were defined to have meaningful impact on the group separation. PLS-DA model was performed via MetaboAnalyst 5.0 (Xia et al., 2009).