Infant Mesenchymal Stem Cell Insulin Action Is Associated With Maternal Plasma Free Fatty Acids, Independent of Obesity Status: The Healthy Start Study

The increased incidence of obesity has led to a parallel rise in type 2 diabetes (1). Alarmingly, data from the SEARCH for Diabetes trial indicate the occurrence of type 2 diabetes in adolescents will quadruple by the year 2050 (2). Although the increased prevalence of Western-style diets and sedentary lifestyles clearly contributes to these trends, it is well accepted that the intrauterine environment has a lasting impact on offspring’s risk of disease later in life. For example, children born to mothers with obesity have an increased risk of obesity (3), type 2 diabetes (4), metabolic syndrome (5), and premature death resulting from cardiovascular events (6). One underlying factor in the pathogenesis of these diseases is impaired insulin action in skeletal muscle and adipose tissues, which contributes to systemic insulin resistance (7).

Offspring in rodent (8,9) and nonhuman primate models (10) of maternal obesity exhibit skeletal muscle insulin resistance. Although much has been gained from studying animal models during pregnancy, discrepancies with humans exist, including a gestational period that is significantly shorter as well as differences in placental physiology, such as estrogen storage/release (11), expression of cell-surface markers (12), and accumulation of diet-specific metabolites (i.e., Neu5gc) (13). Therefore, it is unclear if these preclinical findings can be translated to humans. Our laboratories have used mesenchymal stem cells (MSCs) from the umbilical cord tissue of human offspring donors as an in vitro model of developmental programming in humans, because during fetal development, MSCs differentiate to develop mesodermal tissues, such as skeletal muscle and adipose, and are the primary progenitors for fetal myogenesis as well as postnatal skeletal muscle growth and repair (14–16). We previously reported that MSCs undergoing myogenesis from infants born to mothers with obesity have a greater capacity for adipogenesis and reduced lipid oxidation, both of which are correlated with adiposity of the neonate donor (17,18).

The primary aim of the current study was to determine the effects of maternal obesity on insulin action in human offspring MSCs when undergoing myogenesis in vitro. A secondary aim was to determine the relationships between indices of maternal metabolic health and offspring MSC insulin action. To explore potential mechanisms underlying the relationships between maternal obesity, maternal metabolic health, and offspring MSC insulin action, we also determined the expression of insulin-sensitive markers of cell growth and nutrient storage.

This study used umbilical cord MSCs collected from participants enrolled in the Healthy Start study. Approval for this study was obtained from the University of Colorado Hospital Institutional Review Board, and informed consent was obtained from each participant upon enrollment. All experimental procedures involving MSCs following initial collection and processing were approved and conducted at East Carolina University.

The Healthy Start longitudinal prebirth cohort study recruited women from obstetrics clinics at the University of Colorado Hospital during 2010 to 2014 and enrolled pregnant women age ≥16 years and at ≥23 weeks of gestation, as described in detail elsewhere (19). Briefly, women were excluded if they had prior diabetes, prior premature birth, serious psychiatric illness, or current multiple pregnancy. Pregnant women were evaluated at 27 weeks of gestation for demographics, tobacco use, height, and weight. Fasting blood samples taken at 27 weeks of gestation were drawn for measures of glucose, insulin, triglycerides, and free fatty acids (FFAs). HOMA for insulin resistance (HOMA-IR) was calculated as follows: (fasting serum insulin [mU/L] × fasting plasma glucose [mmol/L])/22.5 (20). Prepregnancy BMI was obtained through medical record abstraction (84%) or self-report (16%) at the first research visit. Umbilical cord–derived MSCs were cultured from a convenience sample of 165 infants collected under the mechanistic arm of Healthy Start: BabyBUMP (18). In this study, a subsample that included mothers with pregravid obesity (n = 10) and normal-weight mothers were frequency matched for maternal age, gestational age at delivery, infant sex, and MSC culture time to confluence (n = 9).

At birth, umbilical cord blood was collected and used for measures of insulin, glucose, triglycerides, and leptin. Infant birth weight was obtained from medical records and measured weight and body composition (fat mass, fat-free mass, and whole-body air displacement plethysmography [PEA POD; COSMED, Inc.]) within 24 to 48 h after birth.

MSCs were cultured from fresh umbilical cord tissue explants of infants born to normal-weight mothers (NW-MSCs) (n = 9) and mothers with obesity (Ob-MSCs) (n = 10), as described previously (18). The MSCs used in the current study were shown to be >98% positive for MSC markers CD73, CD90, and CD105 and were negative for hematopoietic and lymphocyte markers CD34, CD45, and CD19 (18). All experiments were performed on cells in passages 4 to 6. Myogenesis was induced for 20 to 24 days based on visual inspection of myotube formation, which elicited elevated levels of myogenic markers, as described previously (18,21). Day of experimentation did not differ between study groups and did not affect the study results.

Undifferentiated MSCs were incubated with BrdU for 24 h before measurement of BrdU incorporation into newly synthesized DNA using a kit (cat. no. 11647229001; Sigma-Millipore). BrdU incubation time was empirically determined in house for optimal sensitivity. For each cell line, BrdU was measured in triplicate with cell-only controls (no BrdU) and no-cell controls (with BrdU). Triplicate measures were averaged and normalized to cell-only controls.

Following 3 h of serum starvation, myogenically differentiating MSCs were incubated for 2 h with media containing d-[1-¹⁴C] glucose (Perkin-Elmer, Waltham, MA) (1.6 μCi/ml, 5.0 mmol/L glucose), with or without 100 nmol/L insulin, at 37°C (22). Cells were washed twice with ice-cold Dulbecco’s PBS and then solubilized with 0.5% SDS. Lysates were spiked with carrier glycogen (1 mg) and hydrolyzed at 100°C for 1 h. Ice-cold 100% ethanol was added to the spiked lysate to a final concentration of 70% in order to precipitate glycogen. Lysates were centrifuged at 11,100g for 15 min at 4°C and washed with 70% ethanol followed by centrifugation. The glycogen pellets were resuspended with distilled water, and the incorporation of radioactive glucose into glycogen was determined with liquid scintillation. An aliquot of unspiked lysate was reserved to measure protein concentration using bicinchoninic acid assay (Pierce Biotechnology, Rockford, IL). Rates of glycogen synthesis were expressed as either absolute values (nmol glucose/mg protein/minute) to assess basal and maximal glycogen synthesis or log-transformed relative values (log [insulin stimulated/noninsulin stimulated and fold change]) to assess insulin-mediated glycogen synthesis. To clarify interpretation, fold change with insulin was used as the index of insulin action.

Myogenically differentiating offspring MSCs were washed with 1× PBS and harvested in lysis buffer (CelLytic MT; Sigma-Aldrich, St. Louis, MO) supplemented with protease and phosphatase inhibitor cocktails (Sigma-Aldrich). Total protein was determined by bicinchoninic acid assay, and equal total protein from each sample was loaded into Simple Western plates (Protein Simple, San Jose, CA) to measure proteins of interest, as described previously (17,18). Specific content of Akt, phospho-Akt^S473, mTOR, phospho-mTOR^S2448, phospho-p70S6K^T421/S424, phospho-p85S6k^T444/S447, liver x receptor α, sterol regulatory element-binding protein 1c (SREBP-1c), fatty acid synthase, diacylglycerol acyltransferase 1, stearoyl-CoA desaturase 1, glucose transporter 4, acetyl-CoA carboxylase (ACC), and phospho-ACC was measured. Results from Western plates were analyzed using ProteinSimple Compass software. All antibodies were optimized in house for this system; antibody specifics and assay conditions are listed in Supplementary Table 1, and complete chemiluminescent graphs are shown in Supplementary Fig. 5A–N. Each panel of the figure represents the chemiluminescent signals from each participant sample, displayed in an electropherogram format. The electropherogram shows the intensity of the chemiluminescent signatures (y-axis), detected along a spectrum of molecular weights (x-axis), displayed as peaks similar to a band on a Western blot. The area under the curve for the blue or yellow highlighted peak, indicating chemiluminescent intensity for the protein of interest, was used to quantify specific protein content.

Cell pellets were collected from 21-day myogenically differentiating cells. Briefly, cells were rinsed twice with ice-cold PBS, collected, and pelleted by centrifugation. Pellets were flash frozen in liquid nitrogen and stored at −80°C until measurement. For measurement, cells were thawed on ice, resuspended in PBS, and then fortified with internal standards. Lipids were extracted and analyzed by the Colorado Nutrition Obesity Research Center lipidomics core. The core routinely runs triacylglycerol from human biologic samples, and the data are reported for the number of carbons and the number of double bonds. Samples were run on a Sciex 2000 triple quadrupole mass spectrometer (Framingham, MA). Saturated species included only those species in which all three acyl chains contained no double bonds. Concentration was determined by comparing ratios of unknowns to odd chain or deuterated internal standards with standard curves run with standards of each lipid species.

Data were analyzed using GraphPad Prism 9.3 (GraphPad Software; San Jose, CA) or RStudio (version 1.4.1106). All data are presented as means ± SDs. Unpaired Student t tests and one-way repeated measures ANOVAs were used to determine statistical significance, where appropriate. Kolmogorov-Smirnov tests were used to assess normality of the data. If the data did not follow a Gaussian distribution, Mann-Whitney tests were used when indicated in the figure legends. Factors tested were group and/or insulin treatment. Pearson correlations were used to identify relationships between maternal, offspring, and MSC outcomes. Parallel linear regression models were run using the lm function in R, with model adjustment for infant sex, gestational age at delivery, and maternal prepregnancy BMI. Pearson correlations are reported, with adjusted linear model P values additionally reported. Addition of covariates did not appreciably affect results. Statistical significance was set at P ≤ 0.05.

The data sets generated during and/or analyzed during the current study are available from the corresponding author upon reasonable request. No applicable resources were generated or analyzed during the current study.

By design, mothers with obesity had higher BMI but similar maternal age and gestational age at delivery (Table 1). Mothers with obesity had higher fasting plasma insulin concentrations and HOMA-IR values (P ≤ 0.05). Infants from both cohorts were similar in sex distribution, birth weight, and cord blood insulin, glucose, and leptin at the time of delivery (Table 1).

MSC proliferation was not different between groups (Supplementary Fig. 1). Protein expression of myogenic markers, myogenin, and myosin heavy chain was similar between NW-MSCs and Ob-MSCs at days 7 and 21 of differentiation, respectively (Supplementary Fig. 2). Insulin stimulation increased glycogen synthesis rates twofold over basal (Fig. 1A), demonstrating insulin responsiveness. However, neither basal nor maximal rates, nor insulin-mediated rates of glycogen synthesis (log fold change over basal), differed between groups (Fig. 1A and B) (P > 0.05). This log fold change value is hereafter referred to as insulin action. To assess whether this trend was affected by the sex of the offspring, we sex stratified the groups as well and found insulin-mediated glycogen synthesis rates were similar between male and female offspring, regardless of parent BMI classification (Supplementary Fig. 3).

Because we did not observe overt differences in MSC insulin action based on maternal BMI, and given the large degree of variation between participants, we hypothesized that maternal metabolic exposures could influence MSC insulin action and assessed relationships between insulin action and blood indices of maternal metabolic health (glucose, insulin, triglycerides, and FFAs) at 27 weeks of gestation. MSC insulin action was not correlated with maternal glucose (Fig. 2A) (P > 0.05), insulin (Fig. 2B) (P > 0.05), or triglycerides (Fig. 2C) (P > 0.05). Only maternal FFAs displayed a positive correlation with MSC insulin action (Fig. 2D) (r = 0.58; P ≤ 0.05).

Emerging evidence indicates excess lipid availability from the mother may accelerate fetal growth, increasing neonatal adiposity (23–25). Insulin is a key growth factor regulating these processes and is necessary for healthy fetal growth. Therefore, the positive correlation of MSC insulin action with maternal FFAs suggests that insulin action may be linked to growth and lipid storage in offspring. Therefore, we assessed this relationship using birth weight, fat mass at birth, and cord blood leptin as offspring indices of lipid storage. Insulin action in offspring MSCs did not correlate with offspring birth weight or fat mass at birth (Fig. 3A and B) (P > 0.05). However, insulin action showed a strong positive relationship with cord blood leptin (Fig. 3C) (r = 0.69; P ≤ 0.05). To determine whether a similar phenomenon was evident within the cells, we assessed the relationship between MSC insulin action and intracellular triglyceride storage. We found that MSC insulin action was positively correlated with total (Fig. 3D) (r = 0.49; P ≤ 0.05) and positively trending with saturated (Fig. 3E) (r = 0.47; P = 0.07) triglyceride content.

Because observed maternal FFAs were related to MSC insulin action (Fig. 2D), we examined MSC differences based on median stratification of maternal FFAs. Supplementary Table 2 shows maternal and infant characteristics based on this stratification. Mothers in the high FFA group were older at the time of conception and delivered their infants ∼6 days earlier, although this did not translate into differences in infant birth weight, adiposity, or other metabolic measures (Supplementary Table 2). MSCs from the offspring of mothers with higher FFAs displayed a trend toward higher insulin action (P = 0.07) (Supplementary Fig. 4A). Given the observed correlation of MSC insulin action with intracellular triglyceride stores, we also measured protein targets involved with lipid synthesis and storage. Although we observed a trend toward greater liver x receptor α protein content in the high maternal FFA group (P = 0.10) (Fig. 4A), there were no group differences in any downstream proteins (SREBP-1c, fatty acid synthase, or diacylglycerol acyltransferase 1 protein content) (Fig. 4B–D). There were also no differences in glucose transporter 4, ACC (total protein or phosphorylation at serine 79), or stearoyl-CoA desaturase 1, (Supplementary Fig. 4B–D).

Maternal overnutrition enhances placental insulin/IGF/mTOR signaling, which may help partition nutrients to the developing fetus, thereby promoting growth (26,27). Given these fetal growth properties of insulin, we next measured protein markers of the insulin/IGF/mTOR signaling pathway in MSCs from offspring. We observed that MSCs from offspring born to mothers with elevated FFAs during pregnancy also displayed elevated Akt^S473 phosphorylation (Fig. 5A) (P ≤ 0.05) and a trending elevation in mTOR protein content (Fig. 5B) (P = 0.07). Furthermore, protein content of the mTOR substrate, p70S6k^T421/S424 phosphorylation (Fig. 5C) (P ≤ 0.05) was elevated, with a trend toward higher p85S6k^T444/S447 phosphorylation (Fig. 5D) (P = 0.08). These data suggest that elevated maternal FFAs are linked to greater insulin action and insulin/mTOR pathway activation in offspring MSCs.

Animal models and human epidemiologic cohorts indicate that gestational obesity increases the risk of obesity as well as systemic and skeletal muscle insulin resistance in the offspring (3–5,10). However, the pathways by which maternal obesity exposure may induce poor metabolic health in offspring are not well defined, particularly in humans. Although we acknowledge that the MSC model is not a peripheral tissue contributing to infant metabolism at or after birth, recent evidence (17,18,28–30) supports that intrauterine exposures contribute to offspring MSC phenotypes, with relevance to long-term offspring metabolic health. The goal of this study was to determine insulin action in umbilical cord–derived MSCs from human infants exposed to obesity in utero.

Maternal obesity was not associated with differences in MSC insulin action (Fig. 1). Although screening for maternal obesity has provided a valuable prognostic tool for maternal health, metabolites from the mother that cross the placenta, such as glucose and FFAs, likely play a much larger role than maternal body size. For example, Gademan et al. (31) reported a positive relationship between maternal BMI and neonatal adiposity but also noted an independent relationship between maternal lipids and neonatal body composition. Samsuddin et al. (25) observed that gestational weight gain and maternal lipid levels were strongly associated with neonatal adiposity, independent of maternal insulin resistance, blood glucose, or pregravid BMI status. Furthermore, we previously reported that maternal metabolic health indicators, such as maternal glucose, insulin HOMA-IR, FFAs, and HDLs, may be more robust predictors of MSC metabolism than BMI (17,29). In line with these recent findings, we found a relationship between maternal FFAs and MSC insulin action (Fig. 2), which suggests that lipid availability in mid to late gestation, as opposed to maternal obesity status, may be a predictor of MSC insulin action. Importantly, the mothers in the obese group did not present with elevated FFAs, but infants born to mothers with higher FFA levels, regardless of body mass, had MSCs with trending elevations (P = 0.07) in insulin action (Supplementary Fig. 4A) and activation of insulin-mediated growth pathways (Fig. 5).

It may seem counterintuitive that MSCs from infants born to mothers with higher FFAs have elevated insulin action. However, when we consider that excess lipid availability may accelerate fetal growth (23–25,32,33), it is biologically plausible that increased FFA availability is concurrently occurring with enhanced insulin action. In our MSCs, insulin action was positively correlated with both triacylglycerol content in the MSCs and cord blood leptin (Fig. 3C–E), an index of fetal growth (34–37), but newborn body composition did not differ according to maternal FFA concentration (Supplementary Table 2) or MSC insulin action (Fig. 3B). Animal models of adverse intrauterine programming indicate that offspring metabolic consequences, such as insulin resistance, do not manifest until later in life (38–40). In humans, longitudinal analysis shows that positive associations between circulating FFAs and insulin resistance do not develop until adulthood (41). Therefore, it is plausible that the phenotype evident in the MSCs in the current study does not represent newborn body habitus but may reflect predisposition for altered metabolism. With respect to stored triglycerides, these lipids have been shown to be inert in the context of insulin sensitivity when compared with other reactive lipid species (e.g., ceramides, diacylglycerols, and acylcarnitines) (42–45). Therefore, it is possible that elevated triglyceride storage may serve as a means to minimize accumulation of other more reactive species, which would be detrimental to the cell. Taken together, the links between maternal obesity, maternal lipid availability, cord blood leptin, and offspring metabolic health outcomes are likely quite nuanced and require further study.

When mothers were divided into high and low FFA groups, both insulin and mTOR pathway induction in MSCs were elevated in the high FFA group (Fig. 5). Little is known about mTOR signaling in infants or infant MSCs in this context, but abundant research in placental tissue demonstrates that mTOR signaling is elevated with fetal overgrowth (26,27,46) and promotes nutrient transport, oxidative phosphorylation, protein synthesis, and hormone synthesis (47). Moreover, we recently reported that activation of the placental mTORC1 pathway is positively associated with offspring adiposity at birth (30), and placental activation of both mTORC1 and insulin signaling pathways is positively associated with adiposity in childhood (28,30). Recent work by Castillo-Castrejon et al. (48) revealed that insulin sensitivity in placental trophoblasts from mothers with obesity prior to pregnancy and/or who developed gestational diabetes during pregnancy was similar compared with that from their lean counterparts without diabetes. The authors speculated the lack of an obesity effect, similar to that observed in the current study, may serve to ensure adequate nutrient delivery to the fetus (48). Although mTOR and insulin signaling are necessary for healthy fetal growth and tissue development, aberrant hyperactivity may contribute to metabolic dysfunction later in life, as indicated in reports of placental pathway activation being linked to greater child adiposity (28,30).

Despite observing a correlation between MSC insulin action and triacylglycerol stores in the myogenically differentiating MSCs, we did not observe differences in the content of proteins involved in insulin-mediated lipid storage when stratified by maternal FFAs (Fig. 4). mTOR activation can stimulate lipid storage through proteins, such as SREBP (49); however, in the current study, SREBP expression was not different. We recently reported that poor maternal metabolic health, independent of obesity, induces deficits in MSC fat oxidation following myogenic induction (17,29). Indeed, in the absence of upregulated lipid biosynthesis pathways or differences in cellular lipid uptake, we reported that MSCs from infants whose mothers had a less favorable metabolic milieu showed lower fat oxidation and greater triglyceride esterification than those from infants whose mothers had a more favorable metabolic milieu (17,29), suggesting a higher propensity for substrate storage.

Although the current data uncover a potentially interesting role of maternal FFAs in the context of fetal programming, the study was not without limitations. First, the lack of a relationship between maternal FFAs and HOMA-IR and/or BMI status renders it difficult to explain the interparticipant variance in FFAs during pregnancy. Although HOMA-IR is useful in assessing whole-body insulin sensitivity, it is more reflective of hepatic insulin action (50,51). Therefore, future analysis should include dynamic measures of glucose or lipid tolerance examining peripheral tissues that may contribute to maternal FFA concentration. Because our original aim was to investigate the effect of maternal prepregnancy BMI, we selected MSC samples from a subset mothers who had obesity (BMI ≥30 kg/m²) prior to pregnancy and compared them with samples from mothers with normal weight (BMI 18.5–24.9 kg/m²). Therefore, our subsequent correlational and stratified analysis may not be generalizable to a BMI spectrum. Secondly, although our data provide evidence of storage phenotype being invoked by elevated maternal FFAs, we do not have sufficient follow-up data on the children to substantiate this hypothesis. Furthermore, the group with elevated FFAs was still within range of a healthy normal pregnancy. Therefore, future investigations should include a spectrum of maternal BMI, larger sample size, and more longitudinal data to help understand the utility of the MSC model in predicting aspects of offspring health (e.g., metabolic markers and body composition). Additionally, although we assessed the MSCs as they were undergoing myogenesis, we were limited in our ability to extrapolate these findings to the offspring’s mature skeletal muscle. Future analyses should attempt to recapitulate aspects of the intrauterine environment in vitro to determine how this affects the metabolic phenotype of both undifferentiated and myogenically differentiating MSCs alike.

In the current study, maternal obesity was not related to insulin action in offspring-derived MSCs undergoing myogenesis. However, maternal FFA concentrations were positively correlated with MSC insulin action, which seems to be linked to cellular growth and lipid storage, as indicated by activation of mTOR signaling and a positive relationship with intracellular triglyceride content. These data support the importance of the intrauterine environment and extend findings to perturbations in insulin and mTOR signaling pathways in the tissue of offspring. Taken together, the data show MSCs from infants born to mothers with elevated FFAs seem to exhibit upregulation of growth and lipid storage, mediated by activation of insulin and mTOR pathways, which may serve to increase nutrient storage. Furthermore, these data highlight the importance of maternal health in dictating offspring outcomes, independent of BMI status.

Funding. This study was supported by the National Institutes of Health (R01DK117168) (K.E.B.), the American Diabetes Association (1-18-ITCS-016) (K.E.B.), the NIH Environmental Influences on Child Health Outcomes Program (1UG3OD023248) (D.D.), the American Heart Association (15GRNT24470029) (L.E.M.), and a Foundation Doctoral Student Research Grant from the American College of Sports Medicine (19-01128) (A.B.C.). The Healthy Start BabyBUMP Project is supported by grants from the American Heart Association (predoctoral fellowship 14PRE18230008) and by the parent Healthy Start study (R01 DK076648) (D.D.) and the Colorado Clinical and Translational Sciences Institute (UL1 TR001082) for maternal visits and collection of birth measures. The Molecular and Cellular Analytical Core of the Nutrition Obesity Research Center at the University of Colorado Anschutz Medical Campus is supported by the NIH (DK048520).

 This project is solely the responsibility of the authors, and this work does not necessarily represent the official views of the National Institutes of Health.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. A.B.C., J.A.H., and K.E.B. conceived this project. A.B.C., D.Z., J.A.J., B.C.B., Z.W.P., and V.Z. performed experiments and collected and analyzed the data. A.B.C., B.C.B., V.Z., E.M.B., P.K., N.T.B., L.E.M., and K.E.B. interpreted the results. A.B.C., J.A.H., and K.E.B. drafted the manuscript. D.D. conceptualized and implemented the parent Healthy Start study. All authors edited the manuscript and approved the final version. K.E.B. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. These findings have been published in part in abstract form following acceptance and presentation at the 81st Scientific Sessions of the American Diabetes Association, 25–29 June 2021.