All methods except infant indirect calorimetry have been described elsewhere [3,4,5]. The study was approved by the East Carolina University Review Board and informed consent was obtained from each participant upon enrollment. Healthy females with various BMIs, between 18–40 years of age, with singleton pregnancy, without chronic disease, and use of any medication or substances (e.g., SSRI, tobacco) that could affect fetal development were recruited <16 weeks of gestation. Upon enrollment, participants underwent a submaximal modified Balke treadmill test [6] and were assigned their target heart rate (THR) zones that corresponded to maternal HR at 60–80% of maximal oxygen consumption (VO2peak). Two control participants enrolled during COVID-19 and THR was based on the self-reported pre-pregnancy physical activity level and age, using published guidelines [6]. After determining THR zones for each participant, women were randomly assigned (computer-generated randomization, GraphPad Prism Software) to aerobic, resistance, combination (aerobic + resistance) exercise, or a control group; however, for the current study all 3 exercise groups were combined into a single “(ME) group”. Upon randomization, participants exercised according to the American College of Obstetricians and Gynecologists guidelines for the duration of their pregnancy (~16–40 weeks). To ensure that participants performed moderate-intensity exercise (60–80% maximal oxygen consumption and 12–14 rated perceived exertion), every session was supervised, and heart rate was continuously monitored to ensure that participants stayed in their THR zones. The control group performed supervised stretching, breathing, and flexibility exercises at low intensity (<40% VO2peak). ME adherence was calculated by dividing the number of sessions attended by the total number of possible sessions within the participants’ gestational period. Maternal age, parity, pre-pregnancy weight and height and body mass index (BMI, kg/m2), gestational diabetes mellitus status (yes or no), gestational weight gain (GWG), length of gestation, mode of delivery, and breastfeeding status, were abstracted from various sources including pre-screening eligibility and postpartum questionnaires as well as maternal and neonatal electronic health records. At 16 weeks of gestation, we determined maternal BMI, waist-to-hip ratio (3D body scanner, or manually (Gulick tape)), and body fat percentage via skinfold technique and age-adjusted equations [7, 8].

Birth measurements (i.e., weight) and infant sex were extracted from neonatal electronic health records. At 4–6 weeks of age, infant weight, length, BMI, and body fat (%), were measured by trained staff in our pediatric lab. Body fat was calculated using skinfold method [9]. To validate our skinfold measurement, we have obtained infant adiposity using DEXA scan on 8 infants in parallel with obtaining their adiposity using skinfold method. DEXA derived net fat percentage was significantly correlated (r = 0.99, R2 = 98, p < 0.001) with skinfold body fat percentage validating the use of skinfold method for 1-month adiposity measurement for the rest of the cohort. Infant respiration (volume of O2 and CO2) and REE were measured using TrueOne 2400 (PARVO Medics, Sandy, UT, USA) indirect calorimetry system [10, 11]. Indirect calorimetry assessment took place in a thermoneutral room, in low ambient lighting while infant rested, using an infant-specific hood canopy system across ~30 min, or until steady state respiration is reached/obtained. The initial 5–10 min of data was discarded to allow the measures to reach steady state and any infant movement was monitored and later excluded from REE analysis. Infants were “fasted” for a minimum of 1 h prior to the visit.

Unpaired parametric and nonparametric two-tailed t-tests with statistical significance set apriori at p < 0.05 were performed where appropriate using GraphPad Prism version 9.3 (GraphPad Software, San Diego, CA). Variance was similar between groups. Pearson correlations were performed to test any correlations between infant and maternal measurements. ANCOVAs were performed to test group differences while controlling for covariates (i.e., infant sex) using JMP Pro 17 (SAS, Cary, NC).