Identification of early transcriptome signatures in placenta exposed to insulin and obesity

Abstract

OBJECTIVE

The purpose of this study was to investigate the effects of insulin on human placental transcriptome and biological processes in first-trimester pregnancy.

STUDY DESIGN

Maternal plasma and placenta villous tissue were obtained at the time of voluntary termination of pregnancy (7–12 weeks) from 17 lean (body mass index, 20.9 ± 1.5 kg/m²) and 18 obese (body mass index, 33.5 ± 2.6 kg/m²) women. Trophoblast cells were immediately isolated for in vitro treatment with insulin or vehicle. Patterns of global gene expression were analyzed using genome microarray profiling after hybridization to Human Gene 1.1 ST and real time reverse transcription–polymerase chain reaction.

RESULTS

The global trophoblast transcriptome was qualitatively separated in insulin-treated vs untreated trophoblasts of lean women. The number of insulin-sensitive genes detected in the trophoblasts of lean women was 2875 (P < .001). Maternal obesity reduced the number of insulin-sensitive genes recovered by 30-fold. Insulin significantly impaired several gene networks regulating cell cycle and cholesterol homeostasis but did not modify pathways related to glucose transport. Obesity associated with high insulin and insulin resistance, but not maternal hyperinsulinemia alone, impaired the global gene profiling of early gestation placenta, highlighting mitochondrial dysfunction and decreased energy metabolism.

CONCLUSION

We report for the first time that human trophoblast cells are highly sensitive to insulin regulation in early gestation. Maternal obesity associated with insulin resistance programs the placental transcriptome toward refractoriness to insulin with potential adverse consequences for placental structure and function.

Extensive changes in maternal metabolic homeostasis take place over the course of human pregnancy.¹ The development of a physiological state of insulin resistance is a required adaptation to pregnancy. There is a transformation from an insulin sensitive-anabolic phase at early stages to an insulin resistant-catabolic phase in late gestation.^2,3

The metabolic switch is facilitated by sequential changes in the action of insulin. Toward the end of normal pregnancy, all women develop physiological hyperinsulinemia in the basal state, about twice higher compared with pregravid levels.^4–6 The increase in basal plasma insulin concentrations results from a 30% increase in hepatic glucose production to compensate for the decrease in insulin sensitivity in maternal peripheral tissues.^5,6

Most of the maternal adaptations are intensified by maternal obesity, which exaggerates the metabolic abnormalities in response to higher level of insulin resistance.⁷ Pregnancy-specific changes in insulin action are influenced by various hormones and adipokines that are produced by maternal adipose tissue and the placenta.^8–10

The placenta is also a significant source of a variety of cytokines, which link the low-grade inflammation to the insulin resistance of pregnancy.¹¹ As in nonpregnant individuals,¹² tumor necrosis factor alpha impairs the post-receptor insulin signaling cascade in skeletal muscle of pregnant women.¹³ Human placental lactogen has long been cited as a contributor to the decreased insulin sensitivity of pregnancy because of its massive production by the placenta and its increasing concentrations through advancing gestation.^14,15

Although in late gestation a high number of insulin receptors are expressed in placental syncytial membranes, sensitivity of the placenta to maternal insulin has remained controversial.^16–18 The placenta is not considered a classical insulin target tissue with regard to the regulation of glucose transport and utilization. ^19,20 Whereas glucose transporter type 4 insulin-sensitive glucose transporters are expressed in the placenta, they are not being translocated upon insulin stimulation.²¹

Of note, most studies in humans have been performed late in gestation when maternal insulin levels are highest. In contrast to late stages, early pregnancy may be a time of increased placental sensitivity to maternal insulin. Placental size estimated by volume in early pregnancy and by weight at term delivery was strongly related to maternal insulin secretory response in early pregnancy.²² Interestingly, insulin receptors are more abundant on the syncytiotrophoblast in the first trimester placenta compared with the third trimester.²³

We hypothesize that early pregnancy is an optimal period for maternal insulin to regulate insulin sensitive pathways in the placenta. The objective of this study was to examine the response of human placental cells to insulin action in early pregnancy of normal-weight women. Secondarily, obesity was designed as a model to investigate the effects of a high maternal insulin environment in early pregnancy, before pregnancy-induced insulin resistance develops.

Materials and Methods

Study subjects

In this study we define lean as subjects whose pregravid weight for height (body mass index [BMI] kg/m²) was less than 25 kg/m² and obese as subjects whose BMI was greater than 30 kg/m².

This study was approved by the Institutional Review Boards of MetroHealth Medical Center. Volunteers provided written informed consent in accordance with institution guidelines for the protection of human subjects prior to sample collection. Women without medical complications or laboratory signs of infection or history of autoimmune disorders were recruited at the time of voluntary pregnancy termination during the first trimester of pregnancy (weeks 7–12). Placental tissue and maternal blood samples were collected at the time of termination. Anthropometric and metabolic parameters obtained from 33 women are presented in Table 1.

TABLE 1.

Characteristics of the study cohort

Characteristic	Maternal BMI	Maternal age	GA	Insulin, μU/mL	Glucose, mg/dL	HOMA-IR index	Leptin, ng/mL
Lean (BMI <25 kg/m2)	21.2 ± 1.9	25.52 ± 7.0	9.5 ± 2.2	7.0 ± 5.2	75 ± 8	1.3 ± 1.1	16.5 ± 12.8
Obese (BMI >30 kg/m2)	35.2 ± 6.5	24.3 ± 4.9	9.9 ± 2.2	13.2 ± 4.9	77 ± 6	2.1 ± 1.3	40.8 ± 20.8
P value	< .0001	.30	.40	.002	.40	.03	< .001

Data are expressed as means ± SD with 17 lean and 16 obese subjects. Student t test was used for statistical analysis.

BMI, body mass index; GA, gestational age; HOMA-IR, homeostatic assessment model for insulin resistance.

Biological specimen collection

Blood and tissue samples were obtained from 33 women following an 8–10 hour fast. Maternal venous blood samples (10 mL) were collected prior to placement of intravenous lines before the pregnancy termination procedure. The placenta was obtained immediately after the termination procedure. Fresh minced placental villous tissue (~0.5 × 0.5 cm) was washed and digested with trypsin and deoxyribonuclease. Trophoblast cells were purified by density gradient centrifugation as previously described.²⁴ The average yield was 5–8 × 10⁶ cells per gram of tissue, with the cell viability greater than 80%.

Cell culture

Freshly isolated first-trimester trophoblast cells were plated into 12 well plates at a density of 2 × 10⁶ cells/well and cultured overnight in Iscoves’s modified Dulbecco’s modified Eagle’s medium culture medium supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin at 37°C under 5% CO₂. Cells were washed and subsequently serum deprived for 18 hours prior to insulin treatment at a final concentration of 300nM or vehicle for control (saline) for 24 hours. In vitro insulin treatment was performed in trophoblast cells derived from the placenta of obese (n = 6) and lean women (n = 4). Untreated trophoblast cells isolated from first-trimester placenta of obese (n = 5) vs lean women (n = 4) were used as controls and for in vivo gene expression.

RNA extraction and microarray processing

Total ribonucleic acid (RNA) was extracted from untreated and treated placental trophoblasts using an RNeasy kit (QIAGEN Inc, Valencia, CA) and electrophoresed to verify integrity. Purified RNA samples were quantified using a NanoDrop spectrophotometer (Thermo Fisher, Wilmington, DE), and integrity was also assessed by spectrometry (Agilent, Santa Clara, CA). Samples with 28S/18S ratio greater than 1.8 were selected. RNA samples were reversed transcribed into complementary deoxyribonucleic acid (cDNA) using Super- Script first-strand synthesis (Invitrogen, Valencia, CA). Complementary RNA was prepared using 100 ng of each RNA using the SMART cDNA synthesis kit (CLONTECH, Palo Alto, CA).

Gene expression was analyzed via whole-genome microarray profiling after hybridization to Human Gene 1.1 ST Affymetrix arrays (around 36,000 transcripts; Affymetrix, Santa Clara, CA). Affymetrix proprietary software Expression Console was used to convert the fluorescent signals on the scanned images (DAT files) into usable spreadsheets of numerical expression signals. During this conversion signals are quantile normalized using an Expression Console feature called Robust Multichip Analysis.

Quality control (QC) reports were generated and informed us as to the usability of a sample. The positive vs negative area under the curve method was used to select the samples. Samples with values less than 0.8 (an Affymetrix guideline) were dismissed from the study. At that point, expression signals were regenerated using the Robust Multi-chip Analysis and only the QC-passing samples were used.

The average group signal was calculated for each gene. Ratios between group means were calculated. When assessing expression differences between groups, we used the following 2 criteria: (1) the group mean ratio had to be 1.5 or greater and (2) the signals had to be deemed statistically significantly different using the Bayesian Analysis of Microarray data analysis. Expression differences passing these criteria were then presented as the major findings in this study.

Principal component analysis of microarray dataset

All QC-passing samples were then categorized according to group (lean, obese, insulin treated, etc) and were inspected using a principal component analysis (PCA) 3-dimensional plot in the software Partek Genomics Suite version 6.4. Samples portrayed in a plot of the first 3 components could be easily inspected by rotating the plot on screen. Outliers were identified and removed from further study. The groups used for PCA were as follows: (1) insulin-treated cells (n = 4) vs untreated cells (n = 5) from lean women, (2) untreated placental cells of women with high plasma insulin (n = 7) vs low plasma insulin (n = 5), and (3) untreated placental cells of lean (n = 5) vs obese (n = 6).

Network construction and pathway prediction using ClueGO plug-in and Cytoscape

ClueGO plug-in in Cytoscape software (Cytoscape Consortium, San Diego, CA) was used to analyze gene ontology (GO) and functional groups in networks for up- and down-regulated genes. Clue GO is an open-source Java tool that extracts the nonredundant biological information for large clusters of genes, using GO, Kyoto Encyclopedia of Genes and Genomes, and BioCarta. ClueGO is integrated in the latest version of Cytoscape Software version 3.0.2. The degree of connectivity between terms (edges) in the network is calculated using kappa statistics. The significance of the terms and groups is automatically calculated.²⁵

Microarray validation

Validation of the differential gene expression according to microarray results were performed by real-time polymerase chain reaction (RT-PCR) either using a LightCycler FastStart DNA Master SYBR Green I kit for some genes and a Roche dual-probe system using Light Cycler 480 Probes master for others, according to manufacturer’s instructions (Roche, Indianapolis, IN). RNA from insulin-treated and untreated trophoblasts was reversed transcribed using a Superscript II RNase H-reverse transcriptase system (Invitrogen).

Specific primers for each gene were designed according to the microarray probe set for higher reproducibility and are described elsewhere (Appendix; Supplemental Table 1). The table also indicates which primers and probes were used for the dual-probe analysis. Amplifications for each gene were performed in duplicates using 20 ng cDNA samples. Results were normalized for β-actin and expressed as fold changes using the ΔΔcycle threshold method.

Statistics

Values are represented as mean ± SD in Tables 1 and 2. Statistical mean differences were calculated by a Student t test. Statistical significance was set at P <.05.

TABLE 2.

Number of genes regulated by insulin treatment or obesity in trophoblasts derived from lean and obese women

GO term	Associated genes found	P value
Insulin up-regulated genes
Hormones and endocrine pathways	18	9.09E-08
Insulin signaling	41	1.88E-06
Energy metabolism	26	1.28E-09
Cytokine-mediated signaling pathway and inflammatory response	38	8.16E-07
Organic acid and anion transport	46	7.79E-07
Insulin down-regulated
Fatty acid and cholesterol metabolism	50	8.67E-05
Cell cycle regulation	44	2.88E-05
Glycolysis	10	3.08E-06
Obesity regulated genes
Lipid metabolism	38	1.03E-08
Energy and mitochondrial function	35	7.27E-09
Amino acid metabolism	6	2.09E-04

GO term, biological processes identified by the gene ontology analysis (P value for the analysis).

Results

Placental transciptome in early pregnancy

Microarray analysis is an effective way to explore possible mechanisms and give an overall perspective of how the total transcriptome is changed or affected by a determined condition. In this study we used microarray to investigate the effects of in vitro insulin treatment and maternal obesity on global gene expression of primary isolated trophoblasts from first-trimester placentas.

The first approach was to run a PCA in the different data sets used for each comparison. PCA of in vitro response to insulin (24 hours of treatment) in trophoblasts derived from lean women (Figure 1, A) showed a clear separation between treated and untreated cells. Such a cleavage in the response to insulin pattern was not found in vivo when samples were grouped and analyzed based on maternal insulinemia (Figure 1, B).

FIGURE 1. PCA of insulin and obesity effects on total placental transcriptome.

Principal component analysis of insulin and obesity effects on total placental transcriptome. A, Good separation of the samples of untreated cells (blue ) or treated with 300 nM insulin (red ). B, Samples classified according to normal blue (7.0 ± 5.2 μU/mL) and high red (13.2 ± 4.9) plasma insulin levels showing no separation pattern. C, Same samples grouped according to maternal pregravid BMI (blue dots, BMI <25 kg/m², green dots, BMI >30 kg/m²). Each dot represents the total transcriptome per sample analyzed in the arrays.

BMI, body mass index; PCA, principal component analysis.

We classified maternal insulinemia as high when homeostatic assessment model for insulin resistance (HOMA-IR) was higher than 2.0 and low when HOMA-IR was lower than 2.0 (criteria defined using average of HOMA-IR as dividing point). However, when data of the same subjects were analyzed as a function of maternal BMI (Figure 1, C), there was a very defined separation of samples derived from lean and obese women. A PCA analysis of insulin treatment in trophoblasts derived from obese women showed no separation of the groups (data not shown), which is in agreement with the fact that only 87 genes had altered expression, which will be discussed further.

Quantitative sorting of insulin-treated (in vitro) and untreated (in vivo effects of maternal obesity) first-trimester trophoblasts revealed that maternal insulin and obesity modify placental trascriptome. Of 36,000 genes queried, analysis of differentially expressed genes showed that insulin treatment regulated a total of 2875 genes in placenta from lean women (Figure 2). Among these, 1957 were down-regulated and 918 were up-regulated. When trophoblast cells derived from obese women were treated with insulin, the same global pattern of regulation was observed. However, the number of insulin regulated genes was 30 times less than in lean (2875 vs 87).

FIGURE 2. Quantitative sorting of genes differentially regulated in first-trimester placenta.

Quantitative sorting of genes differentially regulated in the placenta of first-trimester pregnancy, with fold changes greater than 1.5 for up-regulated and less than 1.5 for down-regulated genes. Cross-hatched bar indicates the total number of genes in the transcriptome. Second and third bars illustrate genes differentially regulated by in vitro insulin treatment of trophoblast cells derived from first-trimester placenta of obese (n = 6) and lean women (n = 4). Last bar represents gene differentially expressed in first-trimester placenta of obese (n = 5) vs lean women (n = 4). White indicates up-regulated, and black indicates down-regulated.

To extend the observation to a clinical setting, we analyzed placental cells directly derived from obese vs lean women without any further treatment. We observed a differential expression of 1342 genes in obese vs lean women with 90% showing a decreased expression (Figure 2).

Biological processes regulated by insulin

Using Clue GO plugin for Cytoscape software, it was possible to make a functionally grouped annotation network that reflects the relationships between the terms based on the similarity of their associated genes. Network analysis from in vitro insulin-treated cells from lean women is presented in Figure 3, A and B. The size and color of the nodes in the networks reflect the statistical significance of the terms. Figure 3, A shows the relationships between the biological processes from genes up-regulated by insulin treatment.

FIGURE 3. Placental biological processes regulated by insulin treatment of first-trimester trophoblast cells from lean women.

Main placental biological processes regulated by in vitro insulin treatment of first-trimester trophoblast cells from lean women. ClueGO plugin for Cytoscape software (Cytoscape Consortium, San Diego, CA) was used for network and molecular profiling analyses of the microarray data. A, Main biological processes of genes up-regulated by insulin treatment. B, Main biological processes of genes down-regulated by insulin treatment. Significance of the clustering is shown by color code and size of the nodes.

GO, gene ontology.

The main biological processes identified in this network are mainly related to placental endocrine hormone synthesis (chorionic somatomammotropin hormone 1, chorionic gonadotropin, leptin, placental growth factor), cytokine and inflammatory response (colony stimulating factor 1 receptor, chemokine (C-C motif) receptor 1, toll like receptors), and insulin signal transduction (mainly renin-angiotensin system signaling pathways). Cellular transport of anions and organic acids was also up-regulated by insulin treatment in these cells (Table 2).

A network generated with insulin down-regulated genes is illustrated in Figure 3, B. There was a high representation of genes involved in the regulation of cell cycle, fatty acid, and cholesterol metabolism The acyl-CoA synthetase ACSL1, and several genes from the ELO family of fatty acid elongation (ELOV4, ELOV5, ELOV6, and ELOV7) pointed to the regulation of fatty acid–related pathways. ACAT2, which catalyzes the limiting step of cholesterol esterification, and other genes from cholesterol biosynthesis (LSS, CYP7B1, CYP2R1, DHCR24) were down-regulated by insulin. INSIG2, which mediates feedback control of cholesterol synthesis via sterol regulatory element-binding proteins, was also down-regulated by insulin. A summary of the biological processes regulated by insulin and some of the most representative genes involved in each process are listed in Supplemental Table 2 and Table 2, respectively.

Biological processes regulated by maternal obesity

Functional clustering of the data derived from placenta cells of obese women compared with their lean counterparts identified 6 main biological processes described in Figure 4, A. The down-regulated pathways were related to cellular function and structure as well as energy and cell metabolism, and the analysis was first focused on the metabolic cluster (Figure 4, B and Table 2).

FIGURE 4. Processes and clustering of genes in trophoblasts of obese vs lean women.

Main biological processes and functional clustering of genes differentially expressed in trophoblasts of obese vs lean women. A, Global biological pattern after ClueGO-Cytoscape analysis (Cytoscape Consortium, San Diego, CA). B, Functional pattern within the metabolic cluster quantified in panel A.

GO, gene ontology.

Supplemental Table 3 shows lists of regulated genes in each of the metabolic clusters. Several genes encoding for mitochondrial pathways were down-regulated. They include mitochondrial cytochromes (CYP19A1, CYP2R1, CYP27A1, CYP1B1, CYP51A1) and genes involved in cholesterol utilization and steroid synthesis (EBP, HMGCS2, HSD17B12, HSD17B2, HSD17B7, LSS, NSDHL). These data point to an adverse effect of maternal obesity on proper placental mitochondrial function in the first stages of pregnancy. The significant changes in the insulin- and obesity-regulated genes were validated by RT-PCR and are shown in Table 3.

TABLE 3.

Validation of microarray results by RT-PCR analysis

Variable	Microarray FC	Validation FC	P value
Obesity
IL8	−1.5	−1.8	< .05
EBP	−2.2	−2.0	< .05
SC4MOL	−1.8	−2.5	< .05
SCD	−1.8	−3.3	< .05
HSD17B1	−2.5	−2.5	< .05
CYP51A1	−2.1	−1.4	< .05
ACAT2	−2.5	−1.6	< .05
IL1β	3.4	9.7	< .05
Insulin treatment
Leptin	2.0	4.0	< .05
FABP4	1.7	4.2	< .05
PPARγ	2.6	1.6	.10

P values were generated from a Student t test between the ΔCts of the compared groups.

Ct, cycle threshold; FC, fold changes; RT-PCR, real-time polymerase chain reaction.

Comment

The primary finding of this study is the detection of several classic insulin-sensitive pathways in the placenta of first-trimester pregnancy. In vitro treatment of isolated placental cells with insulin induced a number of genes known as downstream mediators of insulin signal transduction such as Ras and protein kinase teta.^26,27 Insulin treatment increased the expression of several sodium-potassium adenosine triphosphatases, ion channels, and transporters as described in muscle and pancreatic beta cells.^28,29

The activation of gene-encoding leptin is in agreement with the insulin regulation in adipose tissue.³⁰ Insulin and its plasma membrane receptor represent a well-conserved biological system critical to cell growth and differentiation. The enhancement of placental specific genes, human chorionic gonadotropin, human placental lactogen, and placental growth hormone suggested that insulin promotes growth and endocrine function of the placenta at early developmental stages.

In contrast, the inhibitory effects of insulin on cell cycle pathways in early gestation placenta are consistent with the resumption of the first meiotic division in oocytes.³¹ The down-regulation of genes related to cell cycle is also in agreement with the action of insulin to block placental endothelial cells in anaphase as a way to minimize cell proliferation.³²

The most important finding is that insulin regulation of placental trophoblast cells involves a strong network of genes implicated in cholesterol metabolism and esterification. Cholesterol is an important lipid component of all cellular membranes. In the placenta, cholesterol needs to be esterified for transport to the fetus and trafficking in and out of cellular organelles. The decreased expression of acetyl-CoA acetyltransferase 2 (ACAT2) by insulin treatment of first-trimester placental cells is an important event in the regulation of cholesterol esterification because ACAT2 catalyzes its limiting step. Decreasing the availability of cholesterol esters may modify membrane composition and formation of lipoprotein particles involved in cholesterol transport.^33,34

Enzymes regulating fatty acid elongation and saturation were also regulated by insulin (Table 2 and Supplemental Table 2). The decreased expression of the insulin-induced gene 2 (INSIG-2) highlights the regulation of placental cholesterol homeostasis.^35,36 These concurrent observations are consistent with recent data showing that changes in cellular cholesterol levels drastically affect cellular functions through the regulation of cellular signaling, exocytosis, and lipid raft composition.³⁷

The classic anabolic effect of insulin on lipid and glucose metabolism was not observed in first-trimester human placenta. In contrast, the type of placental regulation resembles that observed in other nonmetabolic tissues such as the brain or intestine, tissues that both need to transport large amounts of cholesterol.³⁸ Of note, in addition to the down-regulation of a few genes involved in glycolysis (aldolase C, fructose-bisphosphate [ALDOC], phosphofructokinase, platelet [PFKP]), insulin did not stimulate genes involved in glucose transport.

These data, in contrast, do not support data from an older report showing increased 2-deoxyglucose uptake by insulin in cultured placental cells.³⁹ The modest response of glucose pathways to insulin reflect the low sensitivity of glucose response to insulin as demonstrated in term human and rodent placenta.^19,21 Taken together, our data indicate that in vitro insulin treatment activates the in vitro transcription of genes of importance for insulin signaling pathway and lipid homeostasis.

When cells were obtained from the placenta of obese women, they appeared to be 30-fold less sensitive to insulin action than cells from normal-weight women. Interestingly, none of the lipid pathways targeted in lean women were regulated, suggesting that some level of insulin sensitivity was lost when the placenta develops in an obesogenic environment. We speculate that exposure for as short as a few weeks to a high-insulin environment programs a refractory response of insulin-sensitive pathways in early placental trophoblast cells.

This lack of response could also be viewed as a state of desensitization of trophoblast cells to insulin. In other words, once cells have been in contact with high insulin concentrations in vivo, the effect of insulin cannot been rescued, even with pharmacological doses of insulin. Therefore, results resemble the gene expression patterns observed in insulin-resistant states in other models of muscle and adipose cells.^40–46

This global gene profiling approach suggested that maternal obesity creates a unique in utero environment that impairs placental transcriptome in a way that is distinct from the change in maternal insulinemia alone. To further investigate the effect of insulin, we analyzed the placental transcriptome pattern in placenta of normal-weight women with physiological plasma insulin levels (7.0 ± 5.2 μU/mL) and obese women with increased plasma insulin. Contrary to the in vitro data showing a separate pattern of gene expression in the presence and absence of insulin in lean women, PCAs run according to maternal insulin levels showed no differences between groups. However, when analyzed according to maternal BMI, the transcriptome patterns were clearly distinct in the placenta of normal-weight vs obese women (Figure 1).

Hyperinsulinemia is a hallmark of obesity in mammals. During pregnancy, obesity is associated with not only hyperinsulinemia and insulin resistance⁴⁷ but also chronic placental inflammation. ⁴⁸ The dysregulation of insulin response in early pregnancy of obese women may be a cause for the chronic exposure of the growing placenta to higher maternal insulin levels.²² Additionally, the down-regulation of several genes regulating mitochondrial metabolism is in line with recent evidence by our group and others suggesting that obesity induces a state of mitochondrial dysfunction in term human placenta.^49,50

In conclusion, insulin regulates multiple pathways in the placenta in early pregnancy. Our study points to a disruption of cholesterol homeostasis and mitochondrial function as potential adverse outcomes of insulin action in the placenta. We propose that the combination of high maternal insulin levels, insulin resistance, and obesity shape the structure and function of the placenta from early development stages.