Identification and comparative analyses of myocardial miRNAs involved in the fetal response to maternal obesity

Abstract

Human and animal studies show that suboptimal intrauterine environments lead to fetal programming, predisposing offspring to disease in later life. Maternal obesity has been shown to program offspring for cardiovascular disease (CVD), diabetes, and obesity. MicroRNAs (miRNAs) are small, noncoding RNA molecules that act as key regulators of numerous cellular processes. Compelling evidence links miRNAs to the control of cardiac development and etiology of cardiac pathology; however, little is known about their role in the fetal cardiac response to maternal obesity. Our aim was to sequence and profile the cardiac miRNAs that are dysregulated in the hearts of baboon fetuses born to high fat/high fructose-diet (HFD) fed mothers for comparison with fetal hearts from mothers eating a regular diet. Eighty miRNAs were differentially expressed. Of those, 55 miRNAs were upregulated and 25 downregulated with HFD. Twenty-two miRNAs were mapped to human; 14 of these miRNAs were previously reported to be dysregulated in experimental or human CVD. We used an Ingenuity Pathway Analysis to integrate miRNA profiling and bioinformatics predictions to determine miRNA-regulated processes and genes potentially involved in fetal programming. We found a correlation between miRNA expression and putative gene targets involved in developmental disorders and CVD. Cellular death, growth, and proliferation were the most affected cellular functions in response to maternal obesity. Thus, the current study reveals significant alterations in cardiac miRNA expression in the fetus of obese baboons. The epigenetic modifications caused by adverse prenatal environment may represent one of the mechanisms underlying fetal programming of CVD.

early life exposure to either an excess or a deficit in maternal nutrition has been shown to cause modulations in the offspring's body composition and cardiovascular and metabolic function as a result of developmental programming (45). The incidence of maternal overnutrition and obesity is rising rapidly worldwide, and the number of pregnant women who are overweight and obese has increased (83). Children exposed to maternal obesity and gestational diabetes during fetal life have a higher risk of insulin resistance (10), myocardial hypertrophy (117), congenital heart defects (74), and cardiovascular disease (61). Consistent with these human epidemiological data, animal models provide strong evidence that being born to an obese mother increases offspring risk of myocardial dysfunction including ventricular hypertrophy (37) and myocardial fibrosis (51). Although the effects of maternal nutrition on the offspring's epigenetic status are well documented (77), the nature of the fetal molecular pathways modified by maternal obesity remains unknown. There is currently much interest in the use of transcriptional profiling of RNA patterns from potentially affected tissues to identify key regulators involved in initiation and progression of fetal programming.

The recent discovery of microRNAs (miRNAs) has revealed a crucial layer of posttranscriptional gene regulation (73). MiRNAs are a class of small (18–25 nucleotides long) noncoding RNA molecules that posttranscriptionally regulate protein-coding mRNA in both plants and animals. More than 1,000 miRNAs have been identified, many of which are tissue specific and temporally regulated in their expression (38). MiRNAs act as governors of gene expression networks, thereby modifying complex cellular phenotypes in development and pathophysiology. MiRNAs mediate gene silencing by binding to specific target sites within the 3′-untranslated regions (UTR) of mRNA, which will either block the translation or bring about degradation of the transcripts. Many miRNAs are dysregulated in response to cellular stress and can modify essential cellular functions of proliferation, differentiation, and cell death (93). However, the role of miRNAs in fetal programming remains largely unstudied.

The baboon (Papio hamadryas) is a well-characterized nonhuman primate model for biological studies (21, 107). Nonetheless, very little is known about baboon miRNAs. Two recent studies have identified and profiled baboon liver miRNAs that are responsive to dietary fat and cholesterol (56). We and others have studied maternal and fetal baboon physiology in both normal pregnancy and following perturbations that led to developmental programming (65). We have shown that maternal overnutrition in sheep leads to impaired fetal cardiac function and altered insulin signaling (104). Our aim in the present study was to undertake a comprehensive sequencing and profiling analysis of miRNA expression in the hearts of baboon fetuses exposed to maternal overnutrition. We hypothesized that maternal overnutrition combined with high-fat diet and maternal obesity will affect the expression of cardiac miRNA in offspring, potentially changing the expression of key proteins in the heart.

MATERIALS AND METHODS

Animal care and maintenance.

All procedures were approved by the Texas Biomedical Research Institute (Texas Biomed) Institutional Animal Care and Use Committee and conducted in Association for Assessment and Accreditation of Laboratory Animal Care-approved facilities.

System for controlling and recording individual feeding.

The animals were fed between either 0700 and 0900 or 1100 and 1300 as described in detail elsewhere (89). At feeding time, all baboons exited their group cage and passed along a chute and into individual feeding cages. The weight of each baboon was obtained as she crossed an electronic scale system (GSE 665, GSE Scale Systems). The weight recorded was the mean of 50 individual measurements over 3 s. If the standard deviation of the weight measurement was >0.01 of the mean weight, the weight was automatically discarded, and the weighing procedure begun again. Water was continuously available in the feeding cages via individual waterers (Lixit, Napa, CA).

Dietary intervention.

Animals were fed Purina Monkey Diet and Monkey Diet Jumbo (Purina LabDiets, St. Louis, MO). Four months prior to pregnancy, healthy female baboons of similar age and weight were randomly assigned to regular (RD) or high-fat/high-fructose diet (HFD) regimen. The HFD contained 45% energy from fat, 4.62% from glucose, 5.64% from fructose, and 2.32% from sucrose with an energy content of 4.03 kcal/g and free access to a variety of high-fructose sodas. The RD contained 12% energy from fat, 0.29% from glucose, 0.32% from fructose with an energy content of 3.07 kcal/g. The amount of protein and all of the essential minerals and vitamins required for baboon were equalized for the RD and HFD regimens. The dietary intervention was carried through the pregnancy. Pregnant baboons underwent Cesarean sections at 165 days of gestation (0.9 gestation, term 184 days).

Cesarean sections.

Surgical procedures were performed by a fully certified MD or DVM, and postsurgical care was prescribed and monitored by a veterinarian. Cesarean sections were performed at 165 days gestation (0.9 gestation) by standard techniques that have been previously described in detail (22). All baboons were premedicated with ketamine hydrochloride (10 mg/kg im). After intubation, isoflurane (2%) was used to maintain a surgical plane of anesthesia throughout surgery. Following hysterotomy, the umbilical cord was identified and used for fetal exsanguination while the baboons were under general anesthesia as approved by the American Veterinary Medical Association Panel on Euthanasia. The placenta and fetus were removed from the uterus and immediately submitted for morphometric analyses and tissue sampling. Fetal morphometrics were obtained at necropsy. Postoperative maternal analgesia was by buprenorphine hydrochloride (Buprenorphine HCl injection; Hospira, Lake Forest, IL) 0.015 mg/kg/day split as two doses for 3 days. After recovery in individual cages, mothers were returned to their group housing.

Histology.

The excised hearts were washed in PBS, fixed overnight in 4% paraformaldehyde, and embedded in paraffin. After serial sectioning of hearts (apex to base), 7-μm sections were stained with Masson trichrome. Fibrosis areas within sections were measured by visualizing blue-stained areas, exclusive of staining that colocalized with perivascular or intramural vascular structures. Using ImageJ software (http://rsbweb.nih.gov/ij/), we used color-based thresholding to determine blue-stained areas and nonstained myocyte areas from each section. The percentage of total fibrotic area was calculated as the summed blue-stained areas divided by total ventricular area, as described previously (99).

Cardiomyocyte proliferation was assessed in 5 μm paraffin sections with a Ki-67 antibody (Leica Biosystems, Newcastle, UK). Sections were deparaffinized in xylenes, rehydrated through ethanol gradient solutions to PBS, and permeabilized in 0.1% Tween in PBS. Antigen retrieval was achieved using 0.01 M citrate buffer (with 0.1% Triton-X100, Sigma-Aldrich) in a microwave oven for 15 min. Endogenous peroxidase activity was inhibited with 1.5% H₂O₂/methanol for 10 min. The primary Ki-67 antibody was diluted 1:100 prior to use, and the tissue was incubated with the primary antibody overnight at 4°C in a humidified chamber. Detection was performed with the ABC kit (Vector Labs). The sections were counterstained by hematoxylin for 30 s. Negative controls were tissues that were not incubated with the primary antibody. The percentage of proliferating cells was calculated by dividing the number of Ki-67-positive nuclei by total nuclei and multiplying by 100.

RNA isolation.

Overall, this study used 11 fetal hearts: six from baboons born to RD-fed mothers (n = 3 males and 3 females) and five from baboons born to HFD-fed mothers (n = 3 females and 2 males). Total RNA was isolated with the use of RNAeasy kit (Qiagen). The integrity of RNA was tested by spectroscopic analysis and by resolving on denaturing formaldehyde gel.

miRNA sequencing and profiling.

RNA quality control, microRNA sequencing, profiling, and data analysis were performed by LC Sciences (Houston, TX). We sequenced and profiled baboon fetal cardiac miRNAs from RD and HFD groups (Gene Expression Omnibus accession number for the microarray data: GSE43323). Initially, RNA samples pooled from two RD and two HFD hearts (males and females) were processed to generate a cDNA library, which was then used for deep sequencing. The purified cDNA library was used for cluster generation on Illumina's Cluster Station and then sequenced on Illumina GAIIx following vendor's instruction for running the instrument. Raw sequencing reads were obtained using Illumina's Pipeline v1.5 software following sequencing image analysis by Pipeline Firecrest Module and base calling by Pipeline Bustard Module. The extracted sequencing reads were then used in the standard data analysis. Custom-made microarrays were used to profile the expression of fetal cardiac miRNAs in all 11 RNA samples (6 RD and 5 HFD).

RT-PCR.

For validation of the microarray data RT-PCR was performed. We reverse transcribed 5 ng of total RNA using the TaqMan MicroRNA Reverse Transcription Kit. Taqman reactions were conducted with commercially available validated primer/probe sets (Applied Biosystems) and normalized to U18 as internal control. The changes in the threshold cycle (C_T) values were calculated by the equation C_T = C_{T target} − C_{T input}. The fold differences were calculated by 2^−(dCT) method.

Western blot analysis.

Potential miRNA targets were determined by Western blot analysis. The whole cell homogenates were isolated and resolved on 4–20% gradient SDS-polyacrylamide gels. After electrophoresis, the proteins were transferred to nitrocellulose membranes, the membranes were blocked with 5% skim milk in Tween-20/Tris-buffered saline and probed with mouse monoclonal anti-HIF-1α (BD Biosciences), anti-thrombospondin (TSP)-1 (R&D Systems), anti-connective tissue growth factor (CTGF) (Gentex), and anti-p53 (Novus). The blots were then incubated with horseradish peroxidase-conjugated secondary antibodies. Finally, the enhanced chemiluminescence (ECL) reaction was performed and the blots were visualized by G-Box (Syngene). The intensity of each signal spot was transformed into digital data with autobackground subtraction during spot density analysis with the Syngene GeneTools software.

Ingenuity Pathways Analysis.

MiRNA Target Filter Analysis and Core Analysis of miRNA Target Genes were performed by Ingenuity Pathways Analysis (IPA, Ingenuity Systems). This software analysis lists genes in the context of known biological response and regulatory networks as well as other higher-order response pathways. MiRNA targets that were associated with biological functions in the Ingenuity Pathways Knowledge Base were used in the analysis.

Statistics.

All data are expressed as means ± SE. The statistical significance of differences between experimental groups was determined by ANOVA and unpaired two-tailed Student's t-test. P values of <0.05 were considered statistically significant.

RESULTS

General characteristics of experimental animals.

Four months after the beginning of the dietary regimen and prior to conception, dams fed the HFD exhibited increased body weight (Fig. 1A) and LDL-cholesterol (Fig. 1B) and showed a trend for a rise in blood triglycerides compared with RD (Fig. 1D). During pregnancy, dams fed the HFD lost nearly 0.2 kg in body weight, while RD dams gained 1.8 kg (Fig. 1C).

Fig. 1.

Physiological changes in regular diet (RD) and high-fat/high-fructose diet (HFD) mothers prior and during the pregnancy. The HFD increased maternal body weight (A) and blood LDL-cholesterol (B), showing a trend for a rise in blood triglycerides (D) prior to breeding. C: during pregnancy, HFD-fed mothers lost nearly 0.5 kg body wt, while RD-fed mothers gained 1.8 kg. *P < 0.05.

Fetal morphometrics at 165 days of gestation.

Table 1 presents morphometric measures on the placenta and fetuses of RD and HFD dams at cesarean section on day 165. Although there was no difference in placental weight or volume, placental efficiency, expressed as the fetal mass supported per unit placental mass, was reduced in HFD pregnancies compared with control RD (P < 0.02)(Table 1). Fetal body weight was reduced by 16% in HFD pregnancies. Brain and thymus weights were significantly increased in HFD fetuses compared with RD. No change in heart weight was detected (Table 1).

Table 1.

Effect of maternal overnutrition on fetal morphometric parameters at cesarean section (0.9 gestation)

Groups	RD (n = 22)	HFD (n = 5)	P Value
Fetal measures	mean ± SE	mean ± SE
Body weight, g	806 ± 24	675 ± 37	< 0.02*
Fetal measures as percent fetal weight
Placenta	26.0 ± 0.80	32.1 ± 2.50	< 0.01*
Brain	10.4 ± 0.30	11.9 ± 0.60	< 0.02*
Heart	0.61 ± 0.02	0.63 ± 0.04	0.6
Kidneys	0.56 ± 0.02	0.59 ± 0.03	0.5
Liver	2.97 ± 0.06	2.97 ± 0.10	1.0
Lungs	2.16 ± 0.07	2.43 ± 0.10	< 0.09
Pancreas	0.07 ± 0.01	0.07 ± 0.01	0.9
Thymus	0.42 ± 0.02	0.52 ± 0.05	0.048*

RD, regular diet; HFD, high-fat diet.

^*P < 0.05.

Heart histology.

Hematoxylin-and-eosin staining did not reveal any significant differences in myofiber orientation in fetal ventricular tissue from RD and HFD groups (not shown). However, Masson trichrome-stained cardiac sections (Fig. 2) showed increased myocardial fibrosis in HFD fetal hearts (22.05 ± 3.8%, n = 5) compared with RD hearts (2.1 ± 0.76%, n = 6; P = 0.003).

Fig. 2.

Phenotypic characterization of fetal hearts from RD and HFD baboons at the day 165 of gestation. Representative histology (top) and quantification of fibrosis (bottom) in heart sections from RD and HFD baboon fetuses. Masson trichrome staining; n = 5 for each group, *P < 0.05 vs. RD.

MiRNA sequencing in fetal baboon hearts.

We isolated total RNA from the hearts of baboon fetuses. The RNA samples were checked for RNA quality control and processed to generate a cDNA library on which deep sequencing was performed. Overall, 27,857,063 sequence reads were obtained (Table 2). A total of 2,369,584 unique sequences were annotated to annotated small RNA sequences. The small RNA libraries exhibited a diverse size distribution of sequence reads that aligned to the human genome (Fig. 3). miRNAs were the most abundantly expressed small RNAs with 22,611,923 or 81% of total sequences. Other small noncoding RNAs such as small interfering RNAs, small nucleolar RNAs, small nuclear RNAs, and transfer RNAs comprised only 1% (318,109) of the total sequences. On average 19,708,547 sequences (or 87.2%) were mapped to miRBase. We discovered 961 miRNAs in the hearts from baboon fetuses: 601 (68%) were identical to human miRNAs; 23 were mapped to other mammalian miRNAs. Of these mapped miRNAs, 157 showed new genome location, and 180 baboon miRNAs were novel; of these 86 were unmapped to known miRNAs (Table 3).

Table 2.

Summary of standard data analysis results

	SequSeq, n	Mappable SequSeq, %	Unique Seq, n	Mappable Unique Seq, %
Raw	27,857,063		2,369,584
Mapped to mRNA	749,824		14,425
Mapped to other RNAs: rRNA, tRNA, snRNA, snoRNA,	318,109		9,810
Mapped to RepBase	5,537		304
Total mappable for miRNA	22,611,923	100	118,674	100
Mapped to miRBase	19,274,534	85.2	18,955	16
Unmapped to miRBase	2,912,223	12.9	96,136	81.0
Mapped total	19,274,534	85.2	18,955	16.0
No hit	3,337,389	14.8	90,723	84

rRNA, ribosomal RNA; tRNA, transfer RNA; snRNA, small nuclear RNA, snoRNA, small nucleolar RNA. RepBase is a database of prototypic sequences representing repetitive DNA from different eukaryotic species. MirBase is a searchable database of published miRNA sequences and annotation.

Fig. 3.

Length distribution of microRNA (miRNA)-mappable reads.

Table 3.

Known and predicted miRNAs

Known miRNAs	Group	Unique miRs, n
Of Homo sapiens	group 1a	601
Of mammals but novel to Homo sapiens	group 1b	23
Of Homo sapiens and mammals, but with new genome locations	group 1c	157
Predicted miRs
Mapped to known mammal miRs and genome; within hairpins	group 2	17
Mapped to known mammal miRs but unmapped to genome	group 3	77
Unmapped to known miRs but mapped to genome and within hairpin	group 4	86
Total (Unique miRs)		961

microRNA: miRNA or miR.

Dysregulation of fetal cardiac miRNA expression in response to maternal obesity.

MiRNA expression profiling using microarrays is a powerful high-throughput tool capable of monitoring the regulatory networks of the entire genome. To identify regulatory networks involved into the fetal response to maternal obesity, miRNA microarray analysis was performed by LC Sciences. Comprehensive miRNA profiles were generated for 11 fetal hearts from baboons born to RD (n = 3 males and 3 females) or HFD (n = 3 females and 2 males)-fed mothers. The expression pattern of differentially expressed miRNAs is presented as a clustered heat map (Fig. 4). Overall, 80 miRNAs were altered in response to maternal obesity (P < 0.05). Of those, 55 miRNAs were upregulated and 25 downregulated. In total, 22 miRNAs (27.5%) mapped exclusively to human, nine miRNAs mapped to other mammalian species: three miRNAs to Bos taurus (bta-miRs-2436-3p, -1296, -2889), three to Canis familiaris (cfa-miRs-143, -135a-2, -127), one miRNA to Monodelphis domestica (mdo-miR-139), one to Mus musculus (mmu-miR-326), and one to Pongo pygmaeus (ppy-miR-638). The remaining 49 miRNAs were unmapped at the time of analysis. The greatest expression change, >16-fold upregulation, was found for hsa-miR-1296 in HFD hearts compared with RD. Other miRNAs demonstrating notable (>4-fold) nutrient sensitivity included overexpression of hsa-mir-30a, hsa-mir-1-2, hsa-mir-223, hsa-mir-197, hsa-mir-145, hsa-mir-21, and hsa-mir-133a-1. Of the 80 significantly differentially expressed miRNAs, a group of 14 (10 upregulated and 4 downregulated) has previously been reported to be dysregulated in experimental or human cardiovascular diseases. This group includes miR-133a, -197, miR-30a, miR-499, and miR-451 (Table 4). A second group includes differentially expressed miRNA that have not been previously linked to cardiac development and function but have been shown to be involved in other human diseases (Table 5), such as cancer (miR-326, miR-181a, miR-377, miR-584, miR-593, miR-101-1, miR-193, miR-342, miR-99b), diabetes (miR-223, miR-181a), multiple sclerosis (miR-326), and Alzheimer's disease (miR-146).

Fig. 4.

Hierarchical clustering of differentially expressed miRNAs. Heat maps of the known and novel (PCC) miRNA microarray data showing the normalized log2 miRNA expression level for each miRNA (rows) in each sample (columns) for P < 0.001 (top) and P < 0.05 (bottom), according to the indicated color scale. Red indicates overexpression; green represents downregulation.

Table 4.

Differentially expressed fetal cardiac miRNAs previously described as having a functional role in CVDs

miRNA	Role in CVD
Upregulated
hsa-miR-30a	regulates myocardial matrix remodeling (31) and atrial fibrillation (66)
hsa-miR-30c	myocardial infarction (31)
hsa-miR-145	regulates smooth muscle cell plasticity (22)
hsa-miR-451	myocardial infarction (11)
hsa-miR-499	regulates mitochondrial dynamics (105) and myosin isoforms (29)
hsa-miR-223	regulates glut4 expression and cardiomyocyte glucose metabolism (69)
hsa-miR-133a	involved in cardiac hypertrophy (35)
hsa-miR-21	involved in heart failure (100), myocardial infarction (28), and cardiac hypertrophy (97)
hsa-miR-197	involved in cardiac hypertrophy (17)
hsa-miR-139	involved in myocardial infarction (86)
Downregulated
hsa-miR-1-2	involved in cardiomyopathy (113)
hsa-miR-27b	involved in cardiac hypertrophy (113)
hsa-mir-378	targets IGF1R (59), inhibits apoptosis in cardiomyocytes (112)
hsa-miR-18a	regulates myocardial matrix remodeling in age related heart failure (102)

CVD, cardiovascular disease; IGFR1, insulin-like growth factor 1 receptor. The numbers in parentheses are reference list numbers.

Table 5.

Differentially expressed fetal cardiac miRNAs not previously described in the heart

miRNA	Experimentally Observed Function
Upregulated
hsa-miR-326	cancer (26); multiple sclerosis (30)
hsa-miR-483	pancreatitis (9)
hsa-miR-181a	cancer (80); diabetes (58)
hsa-miR-377	nephropathy (106); cancer (72)
hsa-miR-99b	myopathy (32), endometriosis (79)
hsa-miR-584	cancer (49); multiple sclerosis (57)
hsa-miR-146	Alzheimer disease (67)
hsa-miR-593	cancer (54)
hsa-miR-101-1	cancer (53)
Downregulated
hsa-miR-582-3p	osteoarthritis (25)
hsa-miR-505	tumor suppression (110)
hsa-miR-193a	cancer (60)
hsa-miR-342	cancer (48)

The numbers in parentheses refer to reference list numbers.

Quantitative RT-PCR analysis of selected miRNAs.

To confirm the accuracy of the results in the microarray study, we performed real-time PCR on 11 differentially expressed miRNAs. Four criteria were used to select candidate miRNAs: 1) the miRNA must be highly expressed in the heart; 2) the miRNA has to be previously linked to cardiac disease; 3) only one representative from a given miRNA family should be considered; and 4) the miRNA must be a target of a commercially available RT-PCR assay at the time of the work. Table 6 summarizes the data and illustrates the differences in expression between the HFD and RD RNA populations found by RT-PCR. U18 (Applied Biosystems) was used to normalize the RT-PCR data set. The normalized RT-PCR data yielded a correlation of 0.68 (P < 0.02) with microarrays. We found that compared with microarray: 1) the changes in expression of seven miRNAs (63%) were consistent with those determined by microarrays: hsa-mir-30a, hsa-mir-1-2, hsa-mir-223, hsa-mir-197, hsa-mir-18a, hsa-mir-584, hsa-mir-499; 2) changes in the opposite direction to those shown by microarrays were found for two miRNAs: hsa-mir-451, hsa-mir-30c-1, and 3) two miRNAs remained unchanged in contrast to our microarray data: hsa-mir-145, hsa-mir-21.

Table 6.

Summary of differences in miRNA expression between RD and HFD groups for 11 dysregulated miRNAs

miRNAs	Fold Change by Microarray	Fold Change by RT-PCR
Validated
hsa-mir-30a	7.0	3.9*
hsa-mir-1-2	7.0	3.4*
hsa-mir-223	6.0	1.4*
hsa-mir-197	6.0	1.5*
hsa-mir-18a	0.8	0.76*
hsa-mir-584	4	1.4*
hsa-mir-499	1.7	1.3*
Validated with opposite direction
hsa-mir-451	2	0.7*
hsa-mir-30c-1	2	0.8*
Not validated
hsa-mir-145	6.0	1
hsa-mir-21	4.8	1

The microarray data were converted to fold change to directly compare with RT-PCR values, n = 6 RD and 5 HFD,

^*P < 0.05.

Identification of miRNA predicted targets.

miRNAs can regulate a large number of target genes and several databases based on various algorithms are available for predicting the targets of selected miRNAs. Target Scan 5.0, PicTar, and DIANA LAB were used to predict gene targets of the dysregulated miRNAs identified in this study. Overall >1,700 predicted and experimentally observed targets were identified. Using Ingenuity Pathways Analysis IPA), we utilized an miRNA target filter to limit the search to targets expressed in the heart. As a result, >1,500 target genes were identified, 133 of which were experimentally observed and the others classified as “highly predicted.”

A subset of validated and predicted target genes was selected to confirm the expression changes by Western blot analysis. Two extracellular matrix proteins and mediators of cardiac fibrosis in humans and rodents (95, 96), CTGF, and TSP-1 were among the validated targets. TSP-1 appeared to be a target of upregulated miR-1(4) and two downregulated miRNAs: miR-27b (115) and miR-18a (102). We found significantly increased cardiac protein levels of TSP-1 (P < 0.05) in HFD hearts compared with RD hearts (Fig. 5, A and C). Similarly, the protein level of CTGF, a target gene of three upregulated miRNAs, miR-145 (63), miR-133a, and miR-30c (31) and one downregulated miR-18a (102), was significantly higher in HFD vs. RD hearts (P = 0.01) (Fig. 5, A and D). Among the predicted targets, Claudin1 (CLDN1), a tight junction component, was identified as the potential target for four upregulated miRNAs: miR-139, miR-145, miR-584, miR-30a, and Western blot analysis showed 50% reduction in CLDN1 levels in HFD hearts compared with RD (Fig. 5, A and B). Novel miRNAs differentially expressed in HFD hearts compared with RD also had a great number of predicted target genes. Table 7 summarizes the information regarding cardiac-related potential target genes of five upregulated and five downregulated novel miRNAs.

Fig. 5.

Protein expression of potential and validated genes identified by Ingenuity Pathways Analysis (IPA). Western blots were performed on the hearts from baboon fetuses born to RD and HFD mothers, n = 5 for each group. Representative images (A) and quantification bars (B–D) for CLDN1, TSP-1, and CTGF. Images were analyzed using Syngene GeneTools software, data are presented as means ± SE; *P < 0.005. Statistical analysis was done by 1-way ANOVA.

Table 7.

Novel baboon miRNAs differentially expressed in HFD hearts compared with RD

PC#	Fold Change	Seed Region	Targets, n	Sample Target Genes
Upregulated
32270T4	6.3	ACGGGGA	4	PAX2, paired box gene 2; RAP2B, member of RAS oncogene family; PAX2, paired box gene2
2048T4	4.5	AUGAGAG	83	caspase-3; SLC5A6, solute carrier family 5 (sodium-dependent vitamin transporter), member 6; MRPL43, mitochondrial ribosomal protein L43
11444T3	4.3	GGUCCCC	242	CALM1, calmodulin 1; MYH9, myosin, heavy chain 9; ACTN4, actinin, alpha 4, HSPB8, heat shock protein 8
24524T4	2.6	CAGAAUU	289	TP53INP2, tumor protein p53 inducible nuclear protein 2; MAP2, microtubule-associated protein 2, CTGF, connective tissue growth factor
18991T4	2.3	AAGGAUU	127	TXNDC13, thioredoxin domain containing 13; ANK2; ankyrin 2; TNPO1, transportin 1
Downregulated
9971T4	0.8	AAUACAU	656	KLF3, Kruppel-like factor 3; ADAM10, ADAM metallopeptidase domain 10; APP, amyloid beta (A4) precursor protein
2949T3	0.5	AUCCACG	48	ARNTL, aryl hydrocarbon receptor nuclear translocator-like; IGF2BP1, insulin-like growth factor 2 mRNA binding protein 1; HAND2, heart and neural crest derivatives expressed 2
4398T3	0.5	AGCAGCC	432	EIF4E, eukaryotic translation initiation factor 4E; TRAK1, trafficking protein, kinesin binding 1; USP32, ubiquitin specific peptidase 32; CLDN1, claudin1
5663T3	0.4	CAGTCGG	6	AHDC1, AT hook, DNA binding motif, containing 1; FOXJ3, forkhead box J3; GTPBP5, GTP binding protein 5
6274T3	0.4	AACUGGU	145	TPM4, tropomyosin 4; CREB5 cAMP responsive element binding protein 5; GJA1, gap junction protein, alpha 1

The fold change, the 2–8 mer seed region, numbers of predicted target genes according to Targetscan and sample target genes, for each novel miRNA are shown. PC, potential candidate.

IPA of predicted targets.

Using the entire list of identified predicted targets as a starting point, we utilized IPA to reveal potential diseases, molecular functions, physiological systems, and canonical pathways associated with differentially expressed miRNAs. Not surprisingly, the analysis identified developmental disorder and cardiovascular disease as the main diseases associated with maternal overnutrition (324 and 251 molecules, respectively; P < 0.05). Cellular death and survival, growth, and proliferation and cellular development were the most affected molecular and cellular functions in response to maternal overnutrition. We next evaluated changes in cell death and proliferation in the RD and HFD hearts. TUNEL assay showed extremely low levels of cardiomyocyte cell death in ventricular tissue of both RD and HFD fetuses (not shown). In contrast, the proliferation rates measured by Ki-67 staining were significantly higher in HFD hearts compared with RD (P < 0.05, Fig. 6).

Fig. 6.

Increase in cell proliferation in HFD hearts compared with RD. Paraffin tissue sections were stained with anti-Ki-67, a marker of cell proliferation, and counterstained with DAPI. The percentage of proliferating cells was calculated by dividing the number of Ki-67-positive nuclei (*) by total nuclei and multiplying by 100. Representative images from RD and 2 HFD fetuses (A) and quantification bars (B); n = 6 for RD and 5 for HFD group. Images were analyzed using Syngene GeneTools software. Data are presented as means ± SE. *P < 0.05 (1-way ANOVA).

IPA also identified 383 potential transcriptional regulators (not shown). The transcription factors with the highest degree of probability and target molecules were: 1) tumor protein p53 (TP53), 2) peroxisome proliferator-activated receptor gamma (PPAR-γ), and 3) hypoxia-inducible factor 1 alpha (HIF-1α). Western blot analysis showed fourfold decrease in the levels of HIF-1α in HFD hearts compared with RD hearts (P < 0.05, Fig. 7, A and B). The expression of p53 showed a trend toward a decrease, which, however, did not reach a statistical significance (P = 0.1). No differences in PPAR-γ were detected (not shown).

Fig. 7.

Effect of maternal obesity on fetal cardiac expression of transcription factors p53 and HIF-1α. Western blot analysis was performed on the hearts from baboon fetuses born to RD and HFD mothers, n = 5 for each group. Representative images (A) and quantification bars (B). Images were analyzed with Syngene GeneTools software. Data are presented as means ± SE. *P < 0.05 (1-way ANOVA).

DISCUSSION

The incidence of obesity has risen sharply over the past 20 yr and has now reached epidemic proportions, with >1.5 billion adults overweight and 500 million clinically obese adults worldwide (68). In addition to the short-term complications of obesity for both mother and fetus during pregnancy, emerging evidence suggests that maternal obesity has long-term consequences for the health of the offspring (13, 94).

Prenatal and early-life nutrition and stress are among the best documented examples of adverse conditions that predispose the offspring to metabolic and cardiovascular diseases in later life (5). These effects have been confirmed in sheep and rat models of maternal overnutrition (3, 47, 51, 85, 98, 104, 111). The in utero environment can substantially modify how the fetal genome is expressed, thereby exerting stimulatory or inhibitory effects on fetal growth and adiposity.

In this study, we focused on the effect of HFD. We showed a significant increase in maternal body weight and blood LDL-cholesterol in dams fed an obesogenic diet. Despite the fact that HFD-fed mothers did not gain weight during the pregnancy, we found significant physiological changes potentially caused by pregravid maternal obesity. We showed a significant decrease in fetal weight, as well as increase in the weights of brain and thymus in the fetuses of HFD mothers. A reduction in fetal body weights has been shown in nonhuman primate model of maternal obesity (71). In human data, a prospective study of pregnancy outcome in obese women showed that only 13.4% of infants were large for gestational age, while 18.8% of infants were small for gestational age (82). Our recent observations (Maloyan A, Myatt L, unpublished observations) have shown a strongly correlation between birth weight and weight gain during pregnancy. This correlation has been shown before (1). Thus, a 16% reduction in birth weight in our baboon cohort can be at least partially explained by the fact that the HFD mothers lost nearly half a kilogram in body weight during pregnancy. This suggests that cardiac abnormalities seen in the offspring of HFD mothers are due to pre-existing maternal adiposity and cannot be reversed by low weight gain during pregnancy. In contrast to other animal models of maternal obesity (34, 37), the cardiac mass did not change in HFD fetuses. However, we found a significant accumulation of fibrotic tissue in the myocardium of HFD fetuses. Fibrosis is usually a hallmark of aging in various organs including kidney (41), liver (42), pancreas (44), lung (12), and heart (8). Accumulation of myocardial collagen at this early stage of development could lead to progressive increase in ventricular stiffness and impaired cardiac function in offspring. We have previously shown that cardiac function is already impaired by late gestation in a sheep model of maternal obesity (104).

A major role in cardiac fibrosis has been historically attributed to various growth factors, proteolytic enzymes, angiogenic factors, and fibrogenic cytokines (109). However, miRNAs have recently come into focus as regulators of cardiac fibrosis (6). A number of miRNAs have been identified to induce cardiac fibrosis; however, cardiac miR-21 is among the most strongly upregulated in response to variety of cardiac and physiological stresses (62) including the 4.8-fold increase in its levels seen in this study. Further studies are needed to define the role of miR-21 in the cardiac fibrotic remodeling in response to maternal obesity.

Changes in fetal gene expression as result of maternal undernutrition have been previously shown in rat (108), sheep (43), and baboon (77) models. Recent reports from our group have shown that both maternal nutrient reduction and overfeeding in the sheep alters gene transcription in the fetal heart (27, 56) and downregulates fetal skeletal-muscle protein synthesis (116). The mechanisms whereby maternal obesity and nutrient excess in utero increase risk for future metabolic disease are poorly understood but likely include qualitative and quantitative changes in fetal nutrient supply in combination with genetic and epigenetic mechanisms. The in utero environment can substantially modify how the fetal genome is expressed, thereby exerting stimulatory or inhibitory effects on fetal growth and adiposity.

In this study, we have sequenced baboon fetal cardiac miRNA and identified miRNAs that were differentially expressed in response to maternal HFD for the first time. Our initial search has been for differentially expressed miRNAs that have been linked to cardiac development. We show that, for example, downregulation in miR-17-92 cluster has been reported to produce septal defects in mice (103), and miR-181a is involved into cardiac neural crest migration (16). At this point we can speculate that dysregulation of developmentally important miRNAs may explain epidemiological studies that have shown that offspring of obese women are at significantly increased risk for a range of congenital heart defects (14, 74).

Some miRNAs whose expression was affected by maternal HFD were similar to those that are changed in adult cardiac diseases, such cardiac hypertrophy (miR-143, miR-499, and miR-21) (24), heart failure (miR-21 and miR-223) (50), and myocardial infarction (miR-30c, miR-451, and miR-139) (84). There is upregulation of miRNAs involved in enhancing fibrosis (miR-21, miR-499, miRs-133 family, and miRs-30 family) (6) and enhancing intracellular trafficking, and cell adhesion (miR-30 family) (31). Postnatal cardiac function studies are required to determine whether abnormal expression of these miRNAs results into the functional abnormalities in the hearts of HFD offspring later in life. We speculate that the changes we have observed in cardiac miRNA expression will prove detrimental for cardiac function in HFD offspring.

A question facing any researcher with a microarray data set is how much quantitative confidence can be placed in them? RT-PCR is becoming the method of choice in follow-up validation (19) although microarray and RT-PCR data often result in disagreement. In our study, we validated the differential expression of 11 miRNAs with correlation coefficient of 0.68. A survey of the literature reveals widely ranging correlations between microarrays and RT-PCR data of 0.48 to 0.94, illustrating the differences between these technologies (7). Discrepancy could arise from: 1) a distance between the location of the PCR primers and microarray probes; 2) an intensity of array spot with low intensities having considerably lower correlations with RT-PCR data than high intensity spots (7); 3) the difficulties for conventional microarrays to differentiate between members of miRNA families, which often differ by as little as one nucleotide but might exhibit differential expression patterns.

The pathways and biological processes affected by the differentially expressed fetal cardiac miRNAs have yet to be experimentally ascertained. Linking an miRNA to its downstream gene targets is a major challenge in miRNA research. A variety of bioinformatic processes have been developed that predict potential binding sites within the sequence of gene 3′-UTRs in an effort to identify genes regulated by miRNAs. In this study, three algorithms (Target Scan 5.0, PicTar, and DIANA LAB) were used. These are sensitive algorithms with substantial overlap in their predicted targets (92).

Both CTGF and TSP-1 are expressed in the heart during development, whereas their expression is low during normal postnatal life (20, 87). Maternal obesity-mediated upregulation of fetal cardiac CTGF and TSP-1, which regulate extracellular matrix remodeling, suggests that changes in miRNA expression may contribute toward fine tuning of the extracellular matrix proteins potentially leading to fibrosis (20, 88). Claudins regulate many critical developmental processes in vertebrates, with their loss leading to developmental abnormalities or even death (101). We found significant downregulation in CLDN1 protein level in HFD fetal hearts compared with RD. We hypothesize that reduction in CLDN1 might play a role in abnormal fetal cardiac development in response to maternal obesity.

Using IPA, we found that numerous transcriptional regulators, cell signaling molecules, and genes involved in cell death, proliferation, and cardiac development are strongly represented among the potential targets. Interestingly, our data showed a significant increase in cardiomyocyte proliferation in HFD hearts vs. RD. An increase in cell proliferation in the offspring of maternal obesity has been shown before. In rats, maternal HFD increased proliferation of fetal neuroepithelial and neuronal precursor cells (15). In a sheep model of maternal obesity, Ford et al. (39) found an increase in the proliferation of fetal pancreatic β-cell. In a recent study, Eulalio et al. (33) identified a number of miRNAs that had in vitro and in vivo capability to increase cell proliferation in neonatal cardiomyocytes as well as to promote cell cycle re-entry of adult cardiomyocytes. MiRNA(s) responsible for the activation of cardiomyocyte proliferation in our model remains to be elucidated.

Interestingly, some of the largest groups of dysregulated genes in obesity, for example, inflammatory response-related genes, were not represented by the in silico analysis of potential targets. Potential explanations for this observation include that miRNAs may be indirect regulators of inflammatory response genes at the level of transcription control and that our model of maternal HFD in pregnancy does not induce fetal inflammation at 165 days gestation.

It is well known that an evolutionarily conserved orchestra of transcription factors controls cardiac development and function. More recently, the contribution of transcription factors to miRNA expression in the heart has been identified (23). For example, transcription of miR-1, miR-21, miR-206, and miR-133 is directly regulated by serum response factor (SRF) (114); transcription of miR-210 is activated by HIF-1α (46); and transcription of profibrotic miR-21 is regulated by AP-1 (40) and SRF (78). Tight cooperation of three transcription factors, p53, STAT3, and NF-κB, has been shown to regulate miR-21 expression during heart failure (18). In this study, we found significant reduction in the levels of HIF-1α in HFD hearts compared with RD. The involvement of HIF-1α in fetal programming is mostly limited to in utero adaptation to hypoxic conditions (2, 81). It is generally accepted that HIF-1α is stabilized under hypoxia (90). However, prolonged chronic hypoxia, such as seen in obesogenic intrauterine environment (75, 76), has been reported to promote HIF-1α degradation and inhibition of its downstream signaling (70). It has been previously shown that HIF-1α is required for the proper development of the cardiovascular system (55, 91), and reduction in HIF-1α as seen in HFD fetuses might lead to cardiac malformations and abnormal cardiac function during fetal and postnatal life.

In conclusion, maternal HFD before and during pregnancy induces fetal cardiac fibrosis and differential expression of cardiac miRNAs that may contribute to the programming of heart development observed in several animal models. Our study has some limitations. First, because of the sample size of previously archived tissue (three males and two females in the HFD group), we were not able to identify sex-specific changes in gene and miRNA expression. Early studies identified marked sexual dimorphism in the baboon response to high caloric diet with female baboons, overfed as infants, having significantly greater body fat mass, percent of body mass that was fat, and mean fat cell volume compared with females that were underfed or normally fed as infants, while no significant changes were observed in male baboons (64). Second, alteration in the expression of miRNA alone is only a first step in understanding their role in programming and does not prove a functional role. Further experiments using in vitro and in vivo models with genetic manipulation of differentially expressed miRNAs will provide a more complete understanding of their role in fetal heart development in the setting of maternal overnutrition.

GRANTS

This work was supported by grants from Clinical and Translational Science Awards (UL1RR-025767) from the Institute for Integration of Medicine and Science at the University of Texas Health Science Center at San Antonio (to A. Maloyan) and was in part funded by a pilot grant to P. W. Nathanielsz from the Southwest National Primate Research Center base grant (National Institutes of Health P51 RR-33986) and HD-21350.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).

AUTHOR CONTRIBUTIONS

Author contributions: A.M., L.A.C., P.W.N., L.M., and M.J.N. conception and design of research; A.M., S.M., P.W.N., and M.J.N. performed experiments; A.M., S.M., S.H., L.A.C., P.W.N., L.M., and M.J.N. analyzed data; A.M., L.A.C., P.W.N., L.M., and M.J.N. interpreted results of experiments; A.M. and S.M. prepared figures; A.M. drafted manuscript; A.M., L.A.C., P.W.N., L.M., and M.J.N. edited and revised manuscript; A.M., L.A.C., P.W.N., L.M., and M.J.N. approved final version of manuscript.