Gestational diabetes and maternal obesity are associated with sex-specific changes in miRNA and target gene expression in the fetus.

BACKGROUND/
OBJECTIVES: Pregnancies complicated by gestational diabetes (GDM) or maternal obesity have been linked to the development of diabetes, obesity, and fatty liver disease later in life with sex-specific manifestations. Alterations in miRNA expression in offspring exposed to GDM and maternal obesity and effects on hepatic development are unknown. Here, we describe how exposure to maternal obesity in utero leads to sex-specific changes in miRNA and target gene expression in human fetal liver.
METHODS: Candidate miRNA expression was measured in second trimester amniotic fluid (AF) from women with GDM. Targets of differentially expressed miRNAs were determined and pathway enrichment of target genes was performed. MiRNA and target gene expression were measured in a separate cohort of second trimester primary human fetal hepatocytes (PHFH) exposed to maternal obesity via qPCR and western blot. All studies were IRB approved.
RESULTS: GDM-exposed AF had significant increases in miRNAs 199a-3p, 503-5p, and 1268a (fold change (FC) ≥ 1.5, p < 0.05). Female offspring-specific analysis showed enrichment in miRNAs 378a-3p, 885-5p, and 7-1-3p (p < 0.05). MiRNA gene targets were enriched in hepatic pathways. Key genes regulating de novo lipogenesis were upregulated in obesity-exposed PHFH, especially in males. Significantly altered miRNAs in GDM AF were measured in obese-exposed PHFH, with consistent increases in miRNAs 885-5p, 199-3p, 503-5p, 1268a, and 7-1-3p (FC ≥ 1.5, p < 0.05). Female PHFH exposed to maternal obesity had increased expression of miR-885-5p, miR-199-3p, miR-503-5p, miR-1268s, and miR-7-1-3p (p < 0.05), corresponding to decreased target genes expression for ABCA1, PAK4, and INSR. In male PHFHs, no miRNA changes were measured but there was increased expression of ABCA1, PAK4, and INSR (p < 0.05).
CONCLUSIONS: Our data suggest sex-specific changes in miRNA and gene expression in PHFH may be one mechanism contributing to the sexual dimorphism of metabolic disease in offspring exposed to GDM and maternal obesity in utero.

Introduction

Gestational diabetes (GDM) and maternal obesity have profound effects on the intrauterine metabolic milieu, induce marked abnormalities in glucose homeostasis and insulin secretion in the fetus and are linked to obesity, diabetes and non-alcoholic fatty liver disease in the offspring[1-6]. Epidemiologic and animal studies support the concept that there is a critical window of developmental programming during which in utero exposures are associated with an increased susceptibility to obesity and diabetes [7, 8] but the molecular mechanisms underlying this phenomenon are unknown. Alterations in epigenetic modifications, including DNA methylation and histone modifications are proposed as a mechanism by which an in utero exposure can lead to permanent changes in cellular function and ultimately metabolic disease later in life [9].

MicroRNAs comprise a large family of small non-coding RNAs and have emerged as key regulators of metabolic homeostasis, but the role that miRNAs play in fetal development is not well understood. Recent technology has enabled the discovery of circulating miRNAs, which can function as signaling molecules and disease biomarkers. In relation to pregnancy, miRNAs are abundant in maternal plasma, amniotic fluid (AF) and placenta and are involved in the proliferation and differentiation of trophoblast cells and immunological defense[10-13]. Furthermore, abnormalities in miRNA processing are associated with poor placental function and failed embryonic development[11].

Although placental dysfunction is a well-known contributor to fetal growth, little is known about the origin of miRNAs identified in AF or whether placental derived miRNAs enter the fetal compartment. At 16–18 weeks gestational age (GA), AF is comprised of mostly fetal-derived components given that fetal skin is not fully keratinized and fetal urine does not make a substantial contribution until after 20 weeks GA[14]. Therefore it is hypothesized that many of the miRNAs measured in AF are fetal derived and thus alterations in miRNA expression in AF from women with GDM may provide insight into the mechanisms by which GDM affects the developing fetus.

Given the potential role that miRNAs play in fetal development, we investigated whether expression of circulating miRNAs in AF collected at GA 16–18 weeks is altered in women who were later clinically diagnosed with GDM between 24–28 weeks GA. In addition, we sought to determine whether changes in miRNA levels from women with GDM or maternal obesity were associated with abnormal fetal liver development.

Materials and Methods

Study Population

Second trimester amniotic fluid samples

AF samples were collected from women with healthy singleton term pregnancies without maternal health conditions, pregnancy complications, or fetal anomalies. AF specimens were collected from women undergoing amniocentesis at GA 16–18 weeks from 2002–2006 and stored in polypropylene cryogenic vials at −80°C adhering to a strict research protocol[15]. Data abstracted from reproductive genetics charts and post-birth outcome surveys included maternal age, race and ethnicity, GA at amniocentesis, indication for amniocentesis, cytogenetic testing results, sex of offspring and pregnancy outcome data including birth weight, GA at birth, and maternal health history, including complications encountered throughout the duration of the pregnancy. Samples were collected after written informed consent was obtained under a research protocol approved by the University of Pennsylvania and the Children’s Hospital of Philadelphia.

For the present study, we used a nested case control design selecting 20 AF samples from mothers subsequently diagnosed with GDM and 20 control AF samples with no history of maternal GDM. Samples were matched 1:1 for maternal age, gestational age at amniocentesis, maternal race/ethnicity and offspring sex. The most common indication for amniocentesis was advanced maternal age (> 35 years). GDM status was identified through an outcome survey completed one month after delivery and confirmed by measurement of AF c-peptide concentrations >4-fold control as previously reported[16]. Clinical GDM testing is typically performed between 24–28 weeks GA. No data on GDM treatment was collected.

Primary Human Fetal Hepatocytes (PHFH)

Fetal liver was obtained in accordance with an approved institutional review board protocol. Tissue obtained from legally aborted second trimester fetuses between 17–19 weeks gestation from women with BMI> 30 or normal weight women (BMI ≤ 25). Liver was minced in Seglan’s buffer containing 3 mg/mL collagenase I (Roche, Indianapolis, IN) and digested for 30 minutes at 37°C. PHFH cultures were derived following a previously published protocol[17]. Cells were collected by centrifugation, resuspended in William’s media (Gibco, Gaithersburg, MD) and passed through a 21-gauge needle to obtain a single cell suspension. Cells were washed and seeded in 6 well collagen coated plates with Williams media supplemented with nicotinamide (10mmol/L), Hepes (20mmol/L), NaHCO3(17 mmol/L), pyruvate (550 mg/L), ascorbic acid-2-phosphate (0.2 mmol/L), glucose (14 mmol/L), glutamine (2mmol/L) and 10−7 mol/L dexamethasone, ITS+premix containing insulin (6.25 μg/mL), transferrin (6.25 μg/mL) selenious acid (6/25 ng/mL) bovine serum albumin (1.25 ng/mL) and linoleic acid (5.35 μ/mL) antibiotics and 5% fetal bovine serum. After 4 hours the media was changed to remove serum and epidermal growth factor was added (20 ng/mL). All cultures were maintained at 37°C in 5% CO2 with medium changed every 2 days. Experiments were performed at 90% confluence prior to passaging cells.

Multiplex Circulating miRNA Assay

Eighty-eight miRNA candidates were generated from a PubMed literature review with search terms including “miRNA” and “Type I and Type 2 diabetes, GDM, preeclampsia, adipogenesis, obesity, and nonalcoholic fatty liver disease” [13, 18–33] (Supplemental Table 1). Individual candidate miRNA abundance was measured via Multiplex Circulating miRNA assay (Abcam, FirePlex, Cambridge, MA). Samples were digested and hybridized to miRNA specific hydrogel particles with a universal biotinylated adapter labeled with a fluorescent reporter, and quantified with EMD Millipore Guava 8HT flow cytometer. Positive and negative controls were included to reduce inter-plate and inter-well variability. MiRNA spike-in target probes measured hybridization success. Blank hydrogel particles were run to define background fluorescence. Abcam FirePlex Analysis Workbench software was used for data analysis (https://www.abcam.com/kits/multiplex-immunoassays-firefly-analysis-workbench-software). Normalization was performed via geNorm algorithm using the three most stable miRNAs across all samples (hsa-let-7d-5p, hsa-mir-107, and hsa-mir-342-3p) [3435]. Data was log converted to eliminate directional bias. Geometric mean and fold changes were calculated for each miRNA based on normalized expression data.

Prediction of miRNA Targets and Pathway Analysis

TargetScan was used to identify potential gene targets of differentially expressed miRNAs through analysis of seed binding sequence sites [36]. Ingenuity Pathway Analysis (IPA) (Qiagen, Hilden, Germany) identified significantly enriched canonical and toxicological pathways for gene targets of differentially expressed miRNAs.

Total miRNA from PHFH was isolated using miRNeasy (Qiagen). MiRNA expression was measured using qPCR SYBR Green (Qiagen) and normalized with miR-16, selected due to minimal change in CT values across all samples. Fold change was calculated and plotted as mean ± SEM (GraphPad Prism).

Target Gene mRNA and Protein Expression

Total RNA was isolated from PHFH using RNeasy (Qiagen). Complementary DNA was prepared using Invitrogen’s SuperScript IV Vilo master mix and qPCR with Taqman probes for ABCA1, PAK4, and INSR using ACTB and GAPDH as housekeeping genes. Protein was extracted PHFH and quantified using Pierce BCA assay (Thermo Fisher, Waltham, MA). Samples were run on 4–12% bis-tris gels and transferred on nitrocellulose membrane using iBlot2. Antibodies used were PAK4 (1:500, Cell Signaling, Danvers, MA), INSR (Invitrogen, ThermoFisher, 1:1000), β-actin (1:1000, Sigma Aldrich, St. Louis, MO) and secondary antibody (1:10000, Licor, Lincoln, NE).

Statistical Analysis

Two sample t-tests or Mann-Whitney tests were used to compare means between 2 groups. Due to previous studies showing offspring sex as an important variable in fetal programming, separate analyses based on the sex of the offspring were performed. Benjamini Hochberg adjustment for multiple hypothesis testing was applied for the AF circulating miRNA assay, although no candidate miRNAs reached statistical significance using q<0.05. Therefore for the AF circulating miRNA assay, fold change>1.5 or <0.6 and raw p<0.05 was considered statistically significant. For QPCR and western blot studies, p<0.05 was considered significant. SAS 7.0, GraphPad Prism 7.0 and Firefly Analysis Workbench (Abcam, Cambridge, UK) were used to perform the statistical analyses.

Results

Table 1 describes the maternal and infant characteristics of the AF samples used for the circulating miRNA assay. Due to the matching algorithms used for the nested case-control design, there were no differences in maternal race, ethnicity, or GA of amniocentesis, or GA at birth (Table 1). The vast majority of the samples were collected from white, non-Hispanic women. The mean age of all women was 37.4 ± 3.22 years and the most common indication for amniocentesis was due to advanced maternal age. All AF samples studied were collected at a mean GA of 16.2 ± 0.56 weeks and the mean GA at birth was 39.0 ± 1.47 weeks. There were no statistical differences in infant birth weight, indicating that the GDM was well controlled during pregnancy.

Table 1.

Maternal and infant demographics for AF samples. AF: amniotic fluid, GDM: gestational diabetes, GA: gestational age, NS: non-significant.

Characteristic	GDM	Control	P value
Race, N (%)
White	18 (90)	18 (90)	NS
Asian/Pacific Islander	2 (10)	2 (10)	NS
Ethnicity, N (%)
Non-Hispanic	19 (95)	20 (100)	NS
Hispanic	1 (5)	0 (0)	NS
Maternal age
years, mean ± SD	37.4 ± 3.6	37.3 ± 2.8	0.73
GA at amniocentesis	16.2 ± 0.6	16.2 ± 0.5	1
weeks, mean ± SD	16.2 ± 0.6	16.2 ± 0.5	1
GA at birth
weeks, mean ± SD	39.1 ± 1.4	38.9 ± 1.6	0.67
Birth weight
grams, mean ± SD
All offspring	3388.9 ± 525.8	3419.0 ± 479.9	0.83
Female offspring	3078.9 ± 411.4	3345.2 ± 553.2	0.16
Male offspring	3733.4 ± 422.0	3492.8 ± 409.8	0.08

Multiplex Circulating miRNA Assay Results

Eighteen miRNAs had significant changes in expression in GDM AF samples compared to controls with p<0.05 but only 6 also met criteria of FC ≥ 1.5 or ≤ 0.66 (Table 2). In the analysis of all 40 AF samples (20 GDM/20 control), miRNAs hsa-miR-199a-3p, hsa-miR-503-5p, and hsa-miR-1268a were increased in AF of women with GDM (fold change: 1.77, 1.5, 1.77 respectively; FC ≥1.5, p ≤0.05). When the analysis was limited to AF samples from female offspring (n=20; 10 GDM /10 control), GDM samples had significant increases in hsa-miR-378a-3p, hsa-miR-885-5p, and hsa-miR-7-1-3p (FC ≥1.5, p<0.05). When only male offspring samples were analyzed (10 GDM/10 Control), GDM samples again showed a significant increase in hsa-miR-199a-3p (FC ≥1.5, p<0.05).

Table 2.

Differentially expressed miRNAs in GDM exposed Amniotic Fluid

Sample group	miRNA	Fold change	P value	3P-seq tag
All (n=20 pairs)	hsa-mir-138–5p	0.70	0.03	GUGGUCG
	hsa-mir-199a-3p*	1.77	0.01	UGAUGAC
	hsa-mir-15b-5p	1.20	0.03	ACGACGA
	hsa-mir-503–5p*	1.50	0.03	GCGACGA
	hsa-mir-1268a*	1.74	0.03	GUGCGGG
Female (n=10 pairs)	hsa-mir-7–1-3p*	1.62	0.004	AAACAA
	hsa-mir-99a-5p	0.88	0.04	AUGCCCA
	hsa-mir-126–3p1	1.32	0.04	GCCAUGC
	hsa-mir-185–5p	1.35	0.001	AGAGAGG
	hsa-mir-210–3p	0.76	0.007	UGCGUGU
	hsa-mir-486–3p	1.41	0.01	ACGGGG
	hsa-mir-15b-5p	1.31	0.01	ACGACGA
	hsa-mir-378a-3p*	1.50	0.01	UCAGGUC
	hsa-mir-197–3p	1.19	0.02	CACCACU
	hsa-mir-885–5p*	1.64	0.02	CAUUACC
	hsa-mir-146a-5p	1.47	0.03	UCAAGAG
	hsa-mir-302a-3p	1.48	0.03	UCGUGAA
Male (n=10 pairs)	hsa-mir-199a-3p*	1.77	0.01	UGAUGAC

FC ≥ 1.5 or ≤ 0.66 and p<0.05

Pathway Enrichment Results

IPA was performed to determine significantly enriched canonical pathways based on mRNA targets of the differentially expressed miRNAs (Supplemental Tables 2–4). For all samples, pathway analysis identified 161 significantly enriched canonical pathways (p<0.05), 32 for female offspring, and 88 for male offspring. Across all 3 analyses, significantly enriched canonical pathways included cell growth and cell cycle regulation, inflammation, stem cell development and there was significant crossover between enriched canonical pathways for each analysis. Similarly, IPA was used to determine enriched toxicological pathways (Supplemental Table 5) and showed consistent enrichment in liver-based processes including proliferation and apoptosis of hepatocytes and hepatocellular carcinoma pathways.

We did not have access to additional AF cohorts exposed to GDM in order to validate our results of the circulating miRNA assay. Therefore, given the strong enrichment of liver related pathways in the IPA toxicological pathway analysis, we hypothesized that miRNAs enriched in AF of women with GDM may originate in fetal liver and that PHFH exposed to maternal obesity may also be enriched for these miRNA species. Although we did not have BMI data for the women in the GDM cohort, the vast majority of women with advanced maternal age and GDM are obese. Furthermore, the mothers with obesity from which the PHFH were obtained had not completed GDM testing since it is typically performed at 24–28 weeks GA. Although we acknowledge that GDM and maternal obesity are distinct physiological processes, these conditions have considerable overlap in effects on fetal development. In addition, we found that PHFH exposed to maternal obesity had increased expression of key genes regulating de novo lipogenesis, supporting the concept that maternal obesity has profound effects on fetal liver metabolism and the effect was more prominent in PHFC from male offspring (Figure 1). Therefore we measured the differentially expressed miRNAs identified from the GDM exposed AF and their physiologically relevant target gene expression in second trimester PHFH exposed to maternal obesity in an effort to determine whether changes in miRNA expression and effects on target gene expression may be one mechanism by which an altered intrauterine milieu affects fetal development.

Figure 1:

mRNA expression for genes involved in de novo lipogenesis in primary human fetal hepatocytes (PHFH) exposed to maternal obesity. * p<0.05. Data shown as mean ± SEM. A. All samples (n=7–10). B. Male samples (n=4–7). C. Female samples (n=−35). Gray bars = maternal BMI < 25 (control); Black bars = maternal BMI > 30 (obese). mRNA expression normalized to β-actin.

Confirmation of miRNA enrichment and target gene expression in second trimester PHFH exposed to maternal obesity

Clinical characteristics of the source subjects for PHFH samples are shown in Table 3. MiRNAs enriched in GDM AF were measured in second trimester PHFHs exposed to maternal obesity (Figure 2). MiRNA-885-5p expression was significantly increased in maternal obesity exposed PHFH in analyses from all (FC: 2.75, p<0.05) and female only (FC: 3.58, p<0,05) samples (Figure 2a). MiR-199a-3p expression was increased in female PHFH exposed to maternal obesity (FC: 2.02, p<0.05) (Figure 2b) but no significant differences were measured when analyzing all samples together or in male samples. miR-503-5p was increased in female (FC: 2.81, p<0.05) and all PHFH (FC: 1.69, p<0.05) but not in male PHFH (Figure 2c) exposed to maternal obesity. MiR-1268a was increased in only female obesity exposed PHFH (FC: 3.7, p<0.05). miR-7-1-3p was increased in all (FC: 1.75, p<0.05) and female obesity exposed PHFH (FC: 2.63, p<0.05) but not males.

Table 3:

Maternal and fetal demographic data for primary human fetal hepatocyte samples

Characteristic	Maternal Obesity	Control	P Value
Fetal Sex, N(%)	12	10	NS
Male	7 (58)	5 (50)	NS
Female	5 (42)	5 (50)	NS
Maternal BMI
mean ± SD	35.1 ± 2.9	21.6 ± 1.7	<0.001
Maternal Age
mean ± SD	24.3 ± 0.8	23.6 ± 0.7	NS
Gestational Age
weeks, mean ± SD	18.7 ± 0.7	18.7 ± 0.8	NS

Maternal obesity exposed: maternal BMI >30; Control: maternal BMI < 25

Figure 2:

miRNA expression in obese exposed PHFH A. miR-885-5p; B.miR-199-30; C: miR-503-5p; D. miR-1268a; E. miR-7-1-3p. Control: open triangles; Obesity exposed: closed circles. N=7–10 (all); n=3–5 (female) and n=4–7 (male). Exact number of samples per experiment represented on figure.* p<0.05. Data shown as mean ± SEM. A. MiRs normalized to miR-16.

Using TargetScan, we found that several of the de novo lipogenesis genes in Figure 1 are targets of the differentially expressed miRNAs listed in Table 2; FFAR4, PGC1A, SCD and SREBF1 are targets of miR7-3p, SCD is a target of miR-199-3p, and miR885-5p is a target of EHHADH. In the miRNAs and target genes listed above, there is a consistent pattern of increased expression of the miRNA in female PHFH but no change in miRNA expression male PHFH, corresponding to a greater increase in mRNA expression of the target gene in male PHFH compared to female PHFH. Therefore we performed additional experiments to determine if the sex specific changes in miRNA expression corresponded to changes in mRNA expression of additional target genes, selecting ABCA1, PAK4 and INSR as targets of miRNAs 885-5p, 199a-3p and 503-5p respectively. There were no significant changes in ABCA1 mRNA expression in the all or female sample analyses but ABCA1 mRNA expression was significantly increased in male PHFH (FC: 3.64, p<0.05) where no enrichment of miR-885-5p was measured (Figure 3a). ABCA1 is a regulator of reverse cholesterol uptake in the liver and placenta [37]. PAK4 is a target of miR-199a-3p involved in hepatocyte proliferation and growth[38]. No significant changes were measured in PAK4 mRNA expression (Figure 3b) but PAK4 protein levels were significantly decreased in female PHFH exposed to maternal obesity (FC: 0.38, p<0.05) and significantly increased in male samples (FC: 3.16, p<0.05) (Figure 2c). MRNA expression of INSR, a target genes of miR-503-5p and an important regulator of fetal growth and metabolism, was decreased in female PHFH (FC: 0.65, p<0.05) and enriched in male PHFH (FC: 1.86, p<0.05) (Figure 3c). Protein levels of INSR trended higher in both male and female PHFH exposed to maternal obesity but did not reach statistical significance (p=0.08 and p=0.05, respectively) (Figure 3e).

Figure 3:

mRNA and protein expression in obese exposed PHFH. A. ABCA1 mRNA expression normalized. B. PAK4 mRNA expression. C. PAK4 protein levels. D. INSR mRNA expression. E. INSR protein levels. Control: open triangles; Obesity exposed: closed circles. N=7–10 (all); n=3–5 (female) and n=4–7 (male). Exact number of samples per experiment represented on figure.* p<0.05. Data shown as mean ± SEM. mRNA and protein levels normalized to β-actin.

Discussion

To our knowledge, this is the first report of significant changes in miRNA expression in second trimester AF from women who were diagnosed with GDM at 24–28 weeks GA. Although there is limited utility using AF as a biomarker to predict GDM, the enrichment of miRNAs in AF prior to the diagnosis of GDM strongly suggests that altered miRNA expression is a component of the intrauterine milieu in GDM pregnancies and either contributes to maternal metabolic abnormalities or affects the developing fetus much earlier in gestation than clinical GDM testing is currently performed. There is limited information about how miRNAs detected in AF alter fetal development, but several studies measuring miRNAs in serum of women with preeclampsia or diabetes suggest that alterations in miRNAs may impact offspring growth[13, 19, 31, 33].

A second major finding in our study is that miRNA and mRNA expression is dependent on offspring sex in offspring of women with GDM or maternal obesity. Sex specific differences in miRNA expression were noted in both AF and PHFH but the mechanisms contributing to this phenomenon remain unknown. Both epidemiological and animal studies based on the developmental origins of health and disease (DOHaD) hypothesis report that the effects of GDM and maternal obesity on the health of the offspring are sex-specific [39-41]. Work in this field is based on the premise that the gestational period is characterized by rapid cellular differentiation and growth and represents a critical window of exposure during which an altered intrauterine environment characterized by GDM or maternal obesity can lead to permanent changes in cellular function. Here we show that mRNA expression of key gene regulating hepatic de novo lipogenesis are increased to a greater degree in male PHFH exposed to maternal obesity than female obese exposed PHFH. Previous work from our group has shown sex-specific alterations in DNA methylation and gene expression in term placenta exposed to diabetes in pregnancy as well as sex-specific changes in metabolomics analysis from second trimester AF from women subsequently diagnosed with GDM [16, 42]. However, the mechanisms responsible for the sex-specific alterations have not been identified.

Since we did not have access to a separate cohort of second trimester AF samples from women with GDM, and based on the hepatic pathway enrichment of the miRNA target genes from the IPA analysis, we hypothesized that the differentially expressed miRNAs from GDM AF may be derived from or have critical functions in the developing liver. Although we acknowledge that maternal obesity and maternal GDM may have distinct programming effects on the fetus, there is likely much overlap in the maternal phenotypes during second trimester of gestation. Although clinical testing for GDM is not typically performed before GA 24 weeks, we propose that the metabolic pathways affecting fetal development in obese women and those subsequently diagnosed with GDM have significant overlap in the second trimester of gestation.

We identified three miRNAs with significantly increased expression in GDM exposed AF in the analysis of all samples: miR-199a-39, miR-503-5p and miR-1268a. MiR-199a-3p is important to pancreatic beta cell function, hepatocyte growth and differentiation and the development and function of brown and beige adipocytes [18, 22, 43]. In a model of type 2 diabetes induced by a low protein maternal diet, adult offspring with glucose intolerance had increased miR-199a-3p in pancreatic beta cells leading to decreased mTOR signaling [18], an important pathway regulating cell growth and energy metabolism [18]. Inhibiting expression of miR-199a restores mTOR signaling and normalizes insulin secretion [18]. Studies suggest that placental mTOR signaling is an important sensing mechanism of maternal nutrient availability and fetal growth [44, 45]. Increased miR-199a-3p is associated with conditions induced by oxidative stress including steatohepatitis through regulation of the PAK4/MEK/ERK pathway at the protein level [38, 46]. We found that increased miR-199-3p is associated with decreased PAK4 protein expression in female PHFH exposed to maternal obesity but in male obese-exposed PHFH there was increased PAK4 protein expression but no enrichment of miR-199a-3p (Figure 2).

There is limited information describing how miR-503 affects fetal development but gene targets of miR-503 include CCNE1, CDC25A, and CCND2, all of which are involved in cell cycle regulation. Increased concentrations of circulating miR-503 were found in plasma of diabetic patients and in myocardial microvascular endothelial cells and skeletal muscle tissue in an animal model of type 2 diabetes [47-49]. Gene targets of miR-503-5p include FGF2, VEGFA and both IGF1R and INSR due to their sequence homology. Overexpression of miR-503-5p is associated with decreased IGF1R expression [50]. Infants born to mothers with GDM and maternal obesity may have macrosomia in part due to increased fetal insulin production acting as a fetal growth factor and second trimester AF from women diagnosed with GDM have been reported to have a 4–5 fold increase in fetal c-peptide concentrations [16]. For this reason, we hypothesized that exposure to maternal obesity may lead to changes in INSR expression in PHFHs. We found that expression of miR-503-5p was significantly increased and INSR mRNA levels were decreased in female PHFH exposed to maternal obesity, but in male PHFH, there was no change miR-503-5p expression while INSR levels were significantly increased (Figure 3d). However, we found a discrepancy between INSR mRNA expression and INSR protein levels, which were increased in both male and female PHFH exposed to maternal obesity, although this did not reach statistical significance (Figure 3e). It is possible that increased miR-503-5p corresponding to decreased INSR mRNA in female PHFH exposed to maternal obesity could represent a negative feedback response, functioning ultimately to dampen the increased INSR protein expression induced by maternal obesity. However, these relationships are not clear at this time and will require additional experiments in the future.

MiR-1268a was the final miRNA increased in the combined male and female GDM AF [19]. MiR-1268a interacts with SLC8A8, a plasma membrane transporter of creatine, and B4FALNT3, a gene involved in structural modifications of cell-surface N-glycans, which regulate the metabolic function of the developing fetus[19].

When our analyses were limited to AF samples with female offspring, hsa-miR-378a-3p, hsa-miR-7-1-3p and hsa-miR-885-5p had significant increases in expression in GDM exposed AF samples. MiR-378 regulates systemic energy homeostasis and the oxidative capacity of insulin target tissues by repressing carnitine O-acetyltransferase and MED13 [51]. Expression of miR-378a increased in adipose tissue of high fat diet-induced obese mice and during differentiation of preadipocytes [52]. Decreased expression of MAPK1, a target of miR-378a, lead to an induction of adipogenesis [52].

The miR-7 family is an important regulator in fetal pancreas development. In humans, it is the most abundantly expressed miRNA in the pancreatic islet [20]. Nieto et al found that inhibition of miR-7 during early embryogenesis resulted in underdeveloped pancreatic tissue, reduced insulin, and increased glucose intolerance[53]. Additionally inhibition of miR-7 led to increased beta cell death and decreased insulin production[53]. MiR-7 also regulates insulin granule exocytosis from mature pancreatic beta-cells [54, 55].

Several studies describe increased circulating concentrations of miR-885-5p and liver pathologies including fatty liver disease and hepatocellular carcinoma[56] but not much is known about its role in fetal development. MiR-885 is a biomarker for pancreatic cancer, hepatocellular carcinoma, cirrhosis and chronic hepatitis B [57-60]. One of the potential targets of miR-885-5p is ABCA1, a key gene involved in reverse cholesterol transport and phospholipid homeostasis in both liver and placenta [37]. In female PHFH exposed to maternal obesity, we found an enrichment of miR-885-5p expression but no change in ABCA1 mRNA expression, but in male PHFH exposed to maternal obesity there was no change in miR-885-5p expression but a strong increase in ABCA1 expression (Figure 2 and 3). The findings confirm the sex specific findings from miRNAs and target genes above and support the need for additional investigations to fully describe the mechanisms responsible for sex-specific metabolic programming effects.

Our findings are limited by the fact that we only assayed AF at one time period during second trimester of gestation. The GDM AF miRNA enrichment was validated in a separate cohort of second trimester PHFH and the consistent increases in miR-199a-3p, miR-503-5p and miR-885-5p in both cohorts adds confidence to our findings. Since amniocentesis is a rare medical procedure due to a recent switch to cell-free fetal DNA testing for cytogenetic studies, we were unable to validate our AF findings in a separate cohort of AF samples. However, we were able to confirm our findings in a cohort of PHFH obtained from women with obesity who had not yet been tested for GDM. Given the small sample size and the increased maternal age in the AF sample cohort, our findings may not be applicable to other populations of women and their offspring. Additional limitations that may have affected miRNA expression include the inability to study the effect of maternal BMI since these data were not collected and the sex specific trends in GDM offspring birth weight (Table 1). Due to study design and the inherent properties of miRNAs, we were unable to determine if the enriched miRNAs were derived from fetal tissues, placenta or maternal tissues. Additional mechanistic studies are needed to determine not only the origin of the differentially expressed miRNAs but also how they travel to target tissue. Finally, we employed a candidate miRNA approach based on literature review, and may have missed important miRNAs critical to regulating fetal development in the intrauterine environment affected by GDM that may have been detected with an unbiased approach.

In summary, the intrauterine environment associated with GDM and maternal obesity is associated with an increase in miR-885-5p, miR-199a-3p, miR-1268a, miR-7-1-3p and miR-503-5p in second trimester AF and fetal liver exposed to maternal obesity has consistent enrichment of the same miRNAS with a corresponding decrease in expression of target genes ABCA1, PAK4 and INSR. These findings suggest that miRNAs may be a critical component to maternal-placental-fetal communication during gestation and therefore have the potential to affect fetal development.